History, Setup
So for quite some time I have a QNAP TS-873x here, equipped with 8
Western Digital Red 10 TB disks, plus 2 WD Blue 500G M2 SSDs. The QNAP
itself has an AMD Embedded R-Series RX-421MD with 4 cores and was
equipped with 48G RAM.
Initially I had been quite happy, the system is nice. It was fast, it
was easy to get to run and the setup of things I wanted was simple
enough. All in a web interface that tries to imitate a kind of
workstation feeling and also tries to hide that it is actually a
webinterface.
Natually with that amount of disks I had a RAID6 for the disks, plus
RAID1 for the SSDs. And then configured as a big storage pool with the
RAID1 as cache. Below the hood QNAP uses MDADM Raid and LVM (if you
want, with thin provisioning), in some form of emdedded linux. The
interface allows for regular snapshots of your storage with flexible
enough schedules to create them, so it all appears pretty good.
QNAP slow
Fast forward some time and it gets annoying. First off you really
should have regular raid resyncs scheduled, and while you can set
priorities on them and have them low priority, they make the whole
system feel very sluggish, quite annoying. And sure, power failure
(rare, but can happen) means another full resync run. Also, it appears
all of the snapshots are always mounted to some
/mnt/snapshot/something place (df on the system gets quite unusable).
Second, the reboot times. QNAP seems to be affected by the more
features, fuck performance virus, and bloat their OS with more and
more features while completly ignoring the performance. Everytime they
do an upgrade it feels worse. Lately reboot times went up to 10 to
15 minutes - and then it still hadn t started the virtual machines /
docker containers one might run on. Another 5 to 10 minutes for those.
Opening the file explorer - ages on calculating what to show. Trying
to get the storage setup shown? Go get a coffee, but please fetch the
beans directly from the plantation, or you are too fast.
Annoying it was. And no, no broken disks or fan or anything, it all
checks out fine.
Replace QNAPs QTS system
So I started looking around what to do. More RAM may help a little
bit, but I already had 48G, the system itself appears to only do 64G
maximum, so not much chance of it helping enough. Hardware is all fine
and working, so software needs to be changed. Sounds hard, but turns
out, it is not.
TrueNAS
And I found that multiple people replaced the QNAPs own system with a
TrueNAS installation and generally had been happy. Looking further I
found that TrueNAS has a variant called Scale - which is based on
Debian. Doubly good, that, so I went off checking what I may need for
it.
Requirements
Heck, that was a step back. To install TrueNAS you need an HDMI out
and a disk to put it on. The one that QTS uses is too small, so no
option.
QNAPs original internal
USB drive, DOM
So either use one of the SSDs that played cache (and
should do so again in TrueNAS, or get the QNAP original replaced.
HDMI out is simple, get a cheap card and put it into one of the two
PCIe-4x slots, done. The disk thing looked more complicated, as QNAP
uses some internal usb stick thing . Turns out it is just a USB
stick that has an 8+1pin connector. Couldn t find anything nice as
replacement, but hey, there are 9-pin to USB-A adapters.
a 9pin to USB A adapter
With that adapter, one can take some random M2 SSD and an M2-to-USB
case, plus some cabling, and voila, we have a nice system disk.
9pin adapter to USB-A connected with some
more cable
Obviously there isn t a good place to put this SSD case and cable, but the
QNAP case is large enough to find space and use some cable ties to store it
safely. Space enough to get the cable from the side, where the
mainboard is to the place I mounted it, so all fine.
Mounted SSD in its external case
The next best M2 SSD was a Western Digital Red with 500G - and while
this is WAY too much for TrueNAS, it works. And hey, only using a tiny
fraction? Oh so much more cells available internally to use when
others break. Or something
Together with the Asus card mounted I was able to install TrueNAS.
Which is simple, their installer is easy enough to follow, just make
sure to select the right disk to put it on.
Preserving data during the move
Switching from QNAP QTS to TrueNAS Scale means changing from MDADM
Raid with LVM and ext4 on top to ZFS and as such all data on it gets
erased. So a backup first is helpful, and I got myself two external
Seagate USB Disks of 6TB each - enough for the data I wanted to keep.
Copying things all over took ages, especially as the QNAP backup
thingie sucks, it was breaking quite often. Also, for some reason I
did not investigate, the performance of it was real bad. It started at
a maximum of 50MB/s, but the last terabyte of data was copied
at MUCH less than that, and so it took much longer than I anticipated.
Copying back was slow too, but much less so. Of course reading things
usually is faster than writing, with it going around 100MB/s most of
the time, which is quite a bit more - still not what USB3 can actually
do, but I guess the AMD chip doesn t want to go that fast.
TrueNAS experience
The installation went mostly smooth, the only real trouble had been on
my side. Turns out that a bad network cable does NOT help the network
setup, who would have thought. Other than that it is the usual set of
questions you would expect, a reboot, and then some webinterface.
And here the differences start. The whole system boots up much faster.
Not even a third of the time compared to QTS.
One important thing: As TrueNAS scale is Debian based, and hence a
linux kernel, it automatically detects and assembles the old RAID
arrays that QTS put on. Which TrueNAS can do nothing with, so it helps
to manually stop them and wipe the disks.
Afterwards I put ZFS on the disks, with a similar setup to what I had
before. The spinning rust are the data disks in a RAIDZ2 setup, the
two SSDs are added as cache devices. Unlike MDADM, ZFS does not have a
long sync process. Also unlike the MDADM/LVM/EXT4 setup from before,
ZFS works different. It manages the raid thing but it also does the
volume and filesystem parts. Quite different handling, and I m still
getting used to it, so no, I won t write some ZFS introduction now.
Features
The two systems can not be compared completly, they are having a
pretty different target audience. QNAP is more for the user that wants
some network storage that offers a ton of extra features easily
available via a clickable interface. While TrueNAS appears more
oriented to people that want a fast but reliable storage system.
TrueNAS does not offer all the extra bloat the QNAP delivers. Still,
you have the ability to run virtual machines and it seems it comes
with Rancher, so some kubernetes/container ability is there. It lacks
essential features like assigning PCI devices to virtual machines, so
is not useful right now, but I assume that will come in a future
version.
I am still exploring it all, but I like what I have right now. Still
rebuilding my setup to have all shares exported and used again, but
the most important are working already.
tl;dr:
Don t write a Rust linked list library: they are hard to do well, and usually useless.
Use VecDeque, which is great. If you actually need more than VecDeque can do, use one of the handful of libraries that actually offer a significantly more useful API.
If you are writing your own data structure, check if someone has done it already, and consider slotmap or generation_arena, (or maybe Rc/Arc).
Contents
Survey of Rust linked list libraries
I have updated my Survey of Rust linked list libraries.
Background
In 2019 I was writing plag-mangler, a tool for planar graph layout.
I needed a data structure. Naturally I looked for a library to help. I didn t find what I needed, so I wrote rc-dlist-deque. However, on the way I noticed an inordinate number of linked list libraries written in Rust. Most all of these had no real reason for existing. Even the one in the Rust standard library is useless.
Results
Now I have redone the survey. The results are depressing. In 2019 there were 5 libraries which, in my opinion, were largely useless. In late 2022 there are now thirteen linked list libraries that ought probably not ever to be used. And, a further eight libraries for which there are strictly superior alternatives. Many of these have the signs of projects whose authors are otherwise competent: proper documentation, extensive APIs, and so on.
There is one new library which is better for some applications than those available in 2019. (I m referring to generational_token_list, which makes a plausible alternative to dlv-list which I already recommended in 2019.)
Why are there so many poor Rust linked list libraries ?
Linked lists and Rust do not go well together. But (and I m guessing here) I presume many people are taught in programming school that a linked list is a fundamental data structure; people are often even asked to write one as a teaching exercise. This is a bad idea in Rust. Or maybe they ve heard that writing linked lists in Rust is hard and want to prove they can do it.
Double-ended queues
One of the main applications for a linked list in a language like C, is a queue, where you put items in at one end, and take them out at the other. The Rust standard library has a data structure for that, VecDeque.
Five of the available libraries:
Have an API which is a subset of that of VecDeque: basically, pushing and popping elements at the front and back.
Have worse performance for most applications than VecDeque,
Are less mature, less available, less well tested, etc., than VecDeque, simply because VecDeque is in the Rust Standard Library.
For these you could, and should, just use VecDeque instead.
The Cursor concept
A proper linked list lets you identify and hold onto an element in the middle of the list, and cheaply insert and remove elements there.
Rust s ownership and borrowing rules make this awkward. One idea that people have many times reinvented and reimplemented, is to have a Cursor type, derived from the list, which is a reference to an element, and permits insertion and removal there.
Eight libraries have implemented this in the obvious way. However, there is a serious API limitation:
To prevent a cursor being invalidated (e.g. by deletion of the entry it points to) you can t modify the list while the cursor exists. You can only have one cursor (that can be used for modification) at a time.
The practical effect of this is that you cannot retain cursors. You can make and use such a cursor for a particular operation, but you must dispose of it soon. Attempts to do otherwise will see you losing a battle with the borrow checker.
If that s good enough, then you could just use a VecDeque and use array indices instead of the cursors. It s true that deleting or adding elements in the middle involves a lot of copying, but your algorithm is O(n) even with the single-cursor list libraries, because it must first walk the cursor to the desired element.
Formally, I believe any algorithm using these exclusive cursors can be rewritten, in an obvious way, to simply iterate and/or copy from the start or end (as one can do with VecDeque) without changing the headline O() performance characteristics.
IMO the savings available from avoiding extra copies etc. are not worth the additional dependency, unsafe code, and so on, especially as there are other ways of helping with that (e.g. boxing the individual elements).
Even if you don t find that convincing, generational_token_list and dlv_list are strictly superior since they offer a more flexible and convenient API and better performance, and rely on much less unsafe code.
Rustic approaches to pointers-to-and-between-nodes data structures
Most of the time a VecDeque is great. But if you actually want to hold onto (perhaps many) references to the middle of the list, and later modify it through those references, you do need something more. This is a specific case of a general class of problems where the naive approach (use Rust references to the data structure nodes) doesn t work well.
But there is a good solution:
Keep all the nodes in an array (a Vec<Option<T>> or similar) and use the index in the array as your node reference. This is fast, and quite ergonomic, and neatly solves most of the problems. If you are concerned that bare indices might cause confusion, as newly inserted elements would reuse indices, add a per-index generation count.
These approaches have been neatly packaged up in libraries like slab, slotmap, generational-arena and thunderdome. And they have been nicely applied to linked lists by the authors of generational_token_list. and dlv-list.
The alternative for nodey data structures in safe Rust: Rc/Arc
Of course, you can just use Rust s interior mutability and reference counting smart pointers, to directly implement the data structure of your choice.
In many applications, a single-threaded data structure is fine, in which case Rc and Cell/RefCell will let you write safe code, with cheap refcount updates and runtime checks inserted to defend against unexpected aliasing, use-after-free, etc.
I took this approach in rc-dlist-deque, because I wanted each node to be able to be on multiple lists.
Rust s package ecosystem demonstrating software s NIH problem
The Rust ecosystem is full of NIH libraries of all kinds. In my survey, there are: five good options; seven libraries which are plausible, but just not as good as the alternatives; and fourteen others.
There is a whole rant I could have about how the whole software and computing community is pathologically neophilic. Often we seem to actively resist reusing ideas, let alone code; and are ignorant and dismissive of what has gone before. As a result, we keep solving the same problems, badly - making the same mistakes over and over again. In some subfields, working software, or nearly working software, is frequently replaced with something worse, maybe more than once.
One aspect of this is a massive cultural bias towards rewriting rather than reusing, let alone fixing and using.
Many people can come out of a degree, trained to be a programmer, and have no formal training in selecting and evaluating software; this is even though working effectively with computers requires making good use of everyone else s work.
If one isn t taught these skills (when and how to search for prior art, how to choose between dependencies, and so on) one must learn it on the job. The result is usually an ad-hoc and unsystematic approach, often dominated by fashion rather than engineering.
The package naming paradox
The more experienced and competent programmer is aware of all the other options that exist - after all they have evaluated other choices before writing their own library.
So they will call their library something like generational_token_list or vecdeque-stableix.
Whereas the novice straight out of a pre-Rust programming course just thinks what they are doing is the one and only obvious thing (even though it s a poor idea) and hasn t even searched for a previous implementation. So they call their package something obvious like linked list .
As a result, the most obvious names seem to refer to the least useful libraries.
Edited 2022-11-16 23:55 UTC to update numbers of libraries in various categories following updates to the survey (including updates prompted by feedback received after this post first published).
In my tubman setup, I started using ZFS on an old server
I had lying around. The machine is really old though (2011!) and it
"feels" pretty slow. I want to see how much of that is ZFS and how
much is the machine. Synthetic benchmarks show that ZFS may be slower
than mdadm in RAID-10 or RAID-6 configuration, so I want to confirm
that on a live workload: my workstation. Plus, I want easy, regular,
high performance backups (with send/receive snapshots) and there's no
way I'm going to use BTRFS because I find
it too confusing and unreliable.
So off we go.
Installation
Since this is a conversion (and not a new install), our procedure is
slightly different than the official documentation but otherwise
it's pretty much in the same spirit: we're going to use ZFS for
everything, including the root filesystem.
So, install the required packages, on the current system:
root@curie:/home/anarcat# sgdisk -p /dev/sdc
Disk /dev/sdc: 1953525168 sectors, 931.5 GiB
Model: ESD-S1C
Sector size (logical/physical): 512/512 bytes
Disk identifier (GUID): [REDACTED]
Partition table holds up to 128 entries
Main partition table begins at sector 2 and ends at sector 33
First usable sector is 34, last usable sector is 1953525134
Partitions will be aligned on 16-sector boundaries
Total free space is 14 sectors (7.0 KiB)
Number Start (sector) End (sector) Size Code Name
1 48 2047 1000.0 KiB EF02
2 2048 1050623 512.0 MiB EF00
3 1050624 3147775 1024.0 MiB BF01
4 3147776 1953525134 930.0 GiB BF00
Unfortunately, we can't be sure of the sector size here, because the
USB controller is probably lying to us about it. Normally, this
smartctl command should tell us the sector size as well:
root@curie:~# smartctl -i /dev/sdb -qnoserial
smartctl 7.2 2020-12-30 r5155 [x86_64-linux-5.10.0-14-amd64] (local build)
Copyright (C) 2002-20, Bruce Allen, Christian Franke, www.smartmontools.org
=== START OF INFORMATION SECTION ===
Model Family: Western Digital Black Mobile
Device Model: WDC WD10JPLX-00MBPT0
Firmware Version: 01.01H01
User Capacity: 1 000 204 886 016 bytes [1,00 TB]
Sector Sizes: 512 bytes logical, 4096 bytes physical
Rotation Rate: 7200 rpm
Form Factor: 2.5 inches
Device is: In smartctl database [for details use: -P show]
ATA Version is: ATA8-ACS T13/1699-D revision 6
SATA Version is: SATA 3.0, 6.0 Gb/s (current: 6.0 Gb/s)
Local Time is: Tue May 17 13:33:04 2022 EDT
SMART support is: Available - device has SMART capability.
SMART support is: Enabled
Above is the example of the builtin HDD drive. But the SSD device
enclosed in that USB controller doesn't support SMART commands,
so we can't trust that it really has 512 bytes sectors.
This matters because we need to tweak the ashift value
correctly. We're going to go ahead the SSD drive has the common 4KB
settings, which means ashift=12.
Note here that we are not creating a separate partition for
swap. Swap on ZFS volumes (AKA "swap on ZVOL") can trigger lockups and
that issue is still not fixed upstream. Ubuntu recommends using a
separate partition for swap instead. But since this is "just" a
workstation, we're betting that we will not suffer from this problem,
after hearing a report from another Debian developer running this
setup on their workstation successfully.
We do not recommend this setup though. In fact, if I were to redo this
partition scheme, I would probably use LUKS encryption and setup a
dedicated swap partition, as I had problems with ZFS encryption as
well.
Creating pools
ZFS pools are somewhat like "volume groups" if you are familiar with
LVM, except they obviously also do things like RAID-10. (Even though
LVM can technically also do RAID, people typically use mdadm
instead.)
In any case, the guide suggests creating two different pools here:
one, in cleartext, for boot, and a separate, encrypted one, for the
rest. Technically, the boot partition is required because the Grub
bootloader only supports readonly ZFS pools, from what I
understand. But I'm a little out of my depth here and just following
the guide.
Boot pool creation
This creates the boot pool in readonly mode with features that grub
supports:
-O encryption=on -O keylocation=prompt -O keyformat=passphrase:
encryption, prompt for a password, default algorithm is
aes-256-gcm, explicit in the guide, made implicit here
-O acltype=posixacl -O xattr=sa: enable ACLs, with better
performance (not enabled by default)
-O dnodesize=auto: related to extended attributes, less
compatibility with other implementations
-O compression=zstd: enable zstd compression, can be
disabled/enabled by dataset to with zfs set compression=off
rpool/example
-O relatime=on: classic atime optimisation, another that could
be used on a busy server is atime=off
-O canmount=off: do not make the pool mount automatically with
mount -a?
-O mountpoint=/ -R /mnt: mount pool on / in the future, but
/mnt for now
Those settings are all available in zfsprops(8). Other flags are
defined in zpool-create(8). The reasoning behind them is also
explained in the upstream guide and some also in [the Debian
wiki][]. Those flags were actually not used:
-O normalization=formD: normalize file names on comparisons (not
storage), implies utf8only=on, which is a bad idea (and
effectively meant my first sync failed to copy some files,
including this folder from a supysonic checkout). and this
cannot be changed after the filesystem is created. bad, bad, bad.
Side note about single-disk pools
Also note that we're living dangerously here: single-disk ZFS pools
are rumoured to be more dangerous than not running ZFS at all. The
choice quote from this article is:
[...] any error can be detected, but cannot be corrected. This
sounds like an acceptable compromise, but its actually not. The
reason its not is that ZFS' metadata cannot be allowed to be
corrupted. If it is it is likely the zpool will be impossible to
mount (and will probably crash the system once the corruption is
found). So a couple of bad sectors in the right place will mean that
all data on the zpool will be lost. Not some, all. Also there's no
ZFS recovery tools, so you cannot recover any data on the drives.
Compared with (say) ext4, where a single disk error can recovered,
this is pretty bad. But we are ready to live with this with the idea
that we'll have hourly offline snapshots that we can easily recover
from. It's trade-off. Also, we're running this on a NVMe/M.2 drive
which typically just blinks out of existence completely, and doesn't
"bit rot" the way a HDD would.
Also, the FreeBSD handbook quick start doesn't have any warnings
about their first example, which is with a single disk. So I am
reassured at least.
Creating mount points
Next we create the actual filesystems, known as "datasets" which are
the things that get mounted on mountpoint and hold the actual files.
Note that it's unclear to me why those datasets are necessary, but
they seem common practice, also used in this FreeBSD
example. The OpenZFS guide mentions the Solaris upgrades and
Ubuntu's zsys that use that container for upgrades and rollbacks.
This blog post seems to explain a bit the layout behind the
installer.
this creates the actual boot and root filesystems:
Notice here a peculiarity: we must create rpool/var/lib to
create rpool/var/lib/docker otherwise we get this error:
cannot create 'rpool/var/lib/docker': parent does not exist
... and no, just creating /mnt/var/lib doesn't fix that
problem. In fact, it makes things even more confusing because an
existing directory shadows a mountpoint, which is the opposite of
how things normally work.
Also note that you will probably need to change storage driver in
Docker, see the zfs-driver documentation for details but,
basically, I did:
Now that we have everything setup and mounted, let's copy all files
over.
Copying files
This is a list of all the mounted filesystems
for fs in /boot/ /boot/efi/ / /home/; do
echo "syncing $fs to /mnt$fs..." &&
rsync -aSHAXx --info=progress2 --delete $fs /mnt$fs
done
You can check that the list is correct with:
mount -l -t ext4,btrfs,vfat awk ' print $3 '
Note that we skip /srv as it's on a different disk.
On the first run, we had:
root@curie:~# for fs in /boot/ /boot/efi/ / /home/; do
echo "syncing $fs to /mnt$fs..." &&
rsync -aSHAXx --info=progress2 $fs /mnt$fs
done
syncing /boot/ to /mnt/boot/...
0 0% 0.00kB/s 0:00:00 (xfr#0, to-chk=0/299)
syncing /boot/efi/ to /mnt/boot/efi/...
16,831,437 100% 184.14MB/s 0:00:00 (xfr#101, to-chk=0/110)
syncing / to /mnt/...
28,019,293,280 94% 47.63MB/s 0:09:21 (xfr#703710, ir-chk=6748/839220)rsync: [generator] delete_file: rmdir(var/lib/docker) failed: Device or resource busy (16)
could not make way for new symlink: var/lib/docker
34,081,267,990 98% 50.71MB/s 0:10:40 (xfr#736577, to-chk=0/867732)
rsync error: some files/attrs were not transferred (see previous errors) (code 23) at main.c(1333) [sender=3.2.3]
syncing /home/ to /mnt/home/...
rsync: [sender] readlink_stat("/home/anarcat/.fuse") failed: Permission denied (13)
24,456,268,098 98% 68.03MB/s 0:05:42 (xfr#159867, ir-chk=6875/172377)
file has vanished: "/home/anarcat/.cache/mozilla/firefox/s2hwvqbu.quantum/cache2/entries/B3AB0CDA9C4454B3C1197E5A22669DF8EE849D90"
199,762,528,125 93% 74.82MB/s 0:42:26 (xfr#1437846, ir-chk=1018/1983979)rsync: [generator] recv_generator: mkdir "/mnt/home/anarcat/dist/supysonic/tests/assets/\#346" failed: Invalid or incomplete multibyte or wide character (84)
*** Skipping any contents from this failed directory ***
315,384,723,978 96% 76.82MB/s 1:05:15 (xfr#2256473, to-chk=0/2993950)
rsync error: some files/attrs were not transferred (see previous errors) (code 23) at main.c(1333) [sender=3.2.3]
Note the failure to transfer that supysonic file? It turns out they
had a weird filename in their source tree, since then removed,
but still it showed how the utf8only feature might not be such a bad
idea. At this point, the procedure was restarted all the way back to
"Creating pools", after unmounting all ZFS filesystems (umount
/mnt/run /mnt/boot/efi && umount -t zfs -a) and destroying the pool,
which, surprisingly, doesn't require any confirmation (zpool destroy
rpool).
The second run was cleaner:
root@curie:~# for fs in /boot/ /boot/efi/ / /home/; do
echo "syncing $fs to /mnt$fs..." &&
rsync -aSHAXx --info=progress2 --delete $fs /mnt$fs
done
syncing /boot/ to /mnt/boot/...
0 0% 0.00kB/s 0:00:00 (xfr#0, to-chk=0/299)
syncing /boot/efi/ to /mnt/boot/efi/...
0 0% 0.00kB/s 0:00:00 (xfr#0, to-chk=0/110)
syncing / to /mnt/...
28,019,033,070 97% 42.03MB/s 0:10:35 (xfr#703671, ir-chk=1093/833515)rsync: [generator] delete_file: rmdir(var/lib/docker) failed: Device or resource busy (16)
could not make way for new symlink: var/lib/docker
34,081,807,102 98% 44.84MB/s 0:12:04 (xfr#736580, to-chk=0/867723)
rsync error: some files/attrs were not transferred (see previous errors) (code 23) at main.c(1333) [sender=3.2.3]
syncing /home/ to /mnt/home/...
rsync: [sender] readlink_stat("/home/anarcat/.fuse") failed: Permission denied (13)
IO error encountered -- skipping file deletion
24,043,086,450 96% 62.03MB/s 0:06:09 (xfr#151819, ir-chk=15117/172571)
file has vanished: "/home/anarcat/.cache/mozilla/firefox/s2hwvqbu.quantum/cache2/entries/4C1FDBFEA976FF924D062FB990B24B897A77B84B"
315,423,626,507 96% 67.09MB/s 1:14:43 (xfr#2256845, to-chk=0/2994364)
rsync error: some files/attrs were not transferred (see previous errors) (code 23) at main.c(1333) [sender=3.2.3]
Also note the transfer speed: we seem capped at 76MB/s, or
608Mbit/s. This is not as fast as I was expecting: the USB connection
seems to be at around 5Gbps:
anarcat@curie:~$ lsusb -tv head -4
/: Bus 02.Port 1: Dev 1, Class=root_hub, Driver=xhci_hcd/6p, 5000M
ID 1d6b:0003 Linux Foundation 3.0 root hub
__ Port 1: Dev 4, If 0, Class=Mass Storage, Driver=uas, 5000M
ID 0b05:1932 ASUSTek Computer, Inc.
So it shouldn't cap at that speed. It's possible the USB adapter is
failing to give me the full speed though. It's not the M.2 SSD drive
either, as that has a ~500MB/s bandwidth, acccording to its spec.
At this point, we're about ready to do the final configuration. We
drop to single user mode and do the rest of the procedure. That used
to be shutdown now, but it seems like the systemd switch broke that,
so now you can reboot into grub and pick the "recovery"
option. Alternatively, you might try systemctl rescue, as I found
out.
I also wanted to copy the drive over to another new NVMe drive, but
that failed: it looks like the USB controller I have doesn't work with
older, non-NVME drives.
Boot configuration
Now we need to enter the new system to rebuild the boot loader and
initrd and so on.
First, we bind mounts and chroot into the ZFS disk:
mount --rbind /dev /mnt/dev &&
mount --rbind /proc /mnt/proc &&
mount --rbind /sys /mnt/sys &&
chroot /mnt /bin/bash
Next we add an extra service that imports the bpool on boot, to make
sure it survives a zpool.cache destruction:
I had to trim down /etc/fstab and /etc/crypttab to only contain
references to the legacy filesystems (/srv is still BTRFS!).
If we don't already have a tmpfs defined in /etc/fstab:
Strangely, that's not exactly what the author, Jim Salter, did in
his actual test bench used in the ZFS benchmarking
article. The first thing is there's no read test at all, which is
already pretty strange. But also it doesn't include stuff like
dropping caches or repeating results.
So here's my variation, which i called fio-ars-bench.sh for
now. It just batches a bunch of fio tests, one by one, 60 seconds
each. It should take about 12 minutes to run, as there are 3 pair of
tests, read/write, with and without async.
My bias, before building, running and analysing those results is that
ZFS should outperform the traditional stack on writes, but possibly
not on reads. It's also possible it outperforms it on both, because
it's a newer drive. A new test might be possible with a new external
USB drive as well, although I doubt I will find the time to do this.
Results
All tests were done on WD blue SN550 drives, which claims to be
able to push 2400MB/s read and 1750MB/s write. An extra drive was
bought to move the LVM setup from a WDC WDS500G1B0B-00AS40 SSD, a WD
blue M.2 2280 SSD that was at least 5 years old, spec'd at 560MB/s
read, 530MB/s write. Benchmarks were done on the M.2 SSD drive but
discarded so that the drive difference is not a factor in the test.
In practice, I'm going to assume we'll never reach those numbers
because we're not actually NVMe (this is an old workstation!) so the
bottleneck isn't the disk itself. For our purposes, it might still
give us useful results.
Rescue test, LUKS/LVM/ext4
Those tests were performed with everything shutdown, after either
entering the system in rescue mode, or by reaching that target with:
systemctl rescue
The network might have been started before or after the test as well:
systemctl start systemd-networkd
So it should be fairly reliable as basically nothing else is running.
Raw numbers, from the ?job-curie-lvm.log, converted to MiB/s and
manually merged:
test
read I/O
read IOPS
write I/O
write IOPS
rand4k4g1x
39.27
10052
212.15
54310
rand4k4g1x--fsync=1
39.29
10057
2.73
699
rand64k256m16x
1297.00
20751
1068.57
17097
rand64k256m16x--fsync=1
1290.90
20654
353.82
5661
rand1m16g1x
315.15
315
563.77
563
rand1m16g1x--fsync=1
345.88
345
157.01
157
Peaks are at about 20k IOPS and ~1.3GiB/s read, 1GiB/s write in the
64KB blocks with 16 jobs.
Slowest is the random 4k block sync write at an abysmal 3MB/s and 700
IOPS The 1MB read/write tests have lower IOPS, but that is expected.
Rescue test, ZFS
This test was also performed in rescue mode.
Raw numbers, from the ?job-curie-zfs.log, converted to MiB/s and
manually merged:
test
read I/O
read IOPS
write I/O
write IOPS
rand4k4g1x
77.20
19763
27.13
6944
rand4k4g1x--fsync=1
76.16
19495
6.53
1673
rand64k256m16x
1882.40
30118
70.58
1129
rand64k256m16x--fsync=1
1865.13
29842
71.98
1151
rand1m16g1x
921.62
921
102.21
102
rand1m16g1x--fsync=1
908.37
908
64.30
64
Peaks are at 1.8GiB/s read, also in the 64k job like above, but much
faster. The write is, as expected, much slower at 70MiB/s (compared
to 1GiB/s!), but it should be noted the sync write doesn't degrade
performance compared to async writes (although it's still below the
LVM 300MB/s).
Conclusions
Really, ZFS has trouble performing in all write conditions. The
random 4k sync write test is the only place where ZFS outperforms
LVM in writes, and barely (7MiB/s vs 3MiB/s). Everywhere else, writes
are much slower, sometimes by an order of magnitude.
And before some ZFS zealot jumps in talking about the SLOG or some
other cache that could be added to improved performance, I'll remind
you that those numbers are on a bare bones NVMe drive, pretty much as
fast storage as you can find on this machine. Adding another NVMe
drive as a cache probably will not improve write performance here.
Still, those are very different results than the tests performed by
Salter which shows ZFS beating traditional configurations in all
categories but uncached 4k reads (not writes!). That said, those tests
are very different from the tests I performed here, where I test
writes on a single disk, not a RAID array, which might explain the
discrepancy.
Also, note that neither LVM or ZFS manage to reach the 2400MB/s read
and 1750MB/s write performance specification. ZFS does manage to reach
82% of the read performance (1973MB/s) and LVM 64% of the write
performance (1120MB/s). LVM hits 57% of the read performance and ZFS
hits barely 6% of the write performance.
Overall, I'm a bit disappointed in the ZFS write performance here, I
must say. Maybe I need to tweak the record size or some other ZFS
voodoo, but I'll note that I didn't have to do any such configuration
on the other side to kick ZFS in the pants...
Real world experience
This section document not synthetic backups, but actual real world
workloads, comparing before and after I switched my workstation to
ZFS.
Docker performance
I had the feeling that running some git hook (which was firing a
Docker container) was "slower" somehow. It seems that, at runtime, ZFS
backends are significant slower than their overlayfs/ext4 equivalent:
May 16 14:42:52 curie systemd[1]: home-docker-overlay2-17e4d24228decc2d2d493efc401dbfb7ac29739da0e46775e122078d9daf3e87\x2dinit-merged.mount: Succeeded.
May 16 14:42:52 curie systemd[5161]: home-docker-overlay2-17e4d24228decc2d2d493efc401dbfb7ac29739da0e46775e122078d9daf3e87\x2dinit-merged.mount: Succeeded.
May 16 14:42:52 curie systemd[1]: home-docker-overlay2-17e4d24228decc2d2d493efc401dbfb7ac29739da0e46775e122078d9daf3e87-merged.mount: Succeeded.
May 16 14:42:53 curie dockerd[1723]: time="2022-05-16T14:42:53.087219426-04:00" level=info msg="starting signal loop" namespace=moby path=/run/docker/containerd/daemon/io.containerd.runtime.v2.task/moby/af22586fba07014a4d10ab19da10cf280db7a43cad804d6c1e9f2682f12b5f10 pid=151170
May 16 14:42:53 curie systemd[1]: Started libcontainer container af22586fba07014a4d10ab19da10cf280db7a43cad804d6c1e9f2682f12b5f10.
May 16 14:42:54 curie systemd[1]: docker-af22586fba07014a4d10ab19da10cf280db7a43cad804d6c1e9f2682f12b5f10.scope: Succeeded.
May 16 14:42:54 curie dockerd[1723]: time="2022-05-16T14:42:54.047297800-04:00" level=info msg="shim disconnected" id=af22586fba07014a4d10ab19da10cf280db7a43cad804d6c1e9f2682f12b5f10
May 16 14:42:54 curie dockerd[998]: time="2022-05-16T14:42:54.051365015-04:00" level=info msg="ignoring event" container=af22586fba07014a4d10ab19da10cf280db7a43cad804d6c1e9f2682f12b5f10 module=libcontainerd namespace=moby topic=/tasks/delete type="*events.TaskDelete"
May 16 14:42:54 curie systemd[2444]: run-docker-netns-f5453c87c879.mount: Succeeded.
May 16 14:42:54 curie systemd[5161]: run-docker-netns-f5453c87c879.mount: Succeeded.
May 16 14:42:54 curie systemd[2444]: home-docker-overlay2-17e4d24228decc2d2d493efc401dbfb7ac29739da0e46775e122078d9daf3e87-merged.mount: Succeeded.
May 16 14:42:54 curie systemd[5161]: home-docker-overlay2-17e4d24228decc2d2d493efc401dbfb7ac29739da0e46775e122078d9daf3e87-merged.mount: Succeeded.
May 16 14:42:54 curie systemd[1]: run-docker-netns-f5453c87c879.mount: Succeeded.
May 16 14:42:54 curie systemd[1]: home-docker-overlay2-17e4d24228decc2d2d493efc401dbfb7ac29739da0e46775e122078d9daf3e87-merged.mount: Succeeded.
Translating this:
container setup: ~1 second
container runtime: ~1 second
container teardown: ~1 second
total runtime: 2-3 seconds
Obviously, those timestamps are not quite accurate enough to make
precise measurements...
After I switched to ZFS:
mai 30 15:31:39 curie systemd[1]: var-lib-docker-zfs-graph-41ce08fb7a1d3a9c101694b82722f5621c0b4819bd1d9f070933fd1e00543cdf\x2dinit.mount: Succeeded.
mai 30 15:31:39 curie systemd[5287]: var-lib-docker-zfs-graph-41ce08fb7a1d3a9c101694b82722f5621c0b4819bd1d9f070933fd1e00543cdf\x2dinit.mount: Succeeded.
mai 30 15:31:40 curie systemd[1]: var-lib-docker-zfs-graph-41ce08fb7a1d3a9c101694b82722f5621c0b4819bd1d9f070933fd1e00543cdf.mount: Succeeded.
mai 30 15:31:40 curie systemd[5287]: var-lib-docker-zfs-graph-41ce08fb7a1d3a9c101694b82722f5621c0b4819bd1d9f070933fd1e00543cdf.mount: Succeeded.
mai 30 15:31:41 curie dockerd[3199]: time="2022-05-30T15:31:41.551403693-04:00" level=info msg="starting signal loop" namespace=moby path=/run/docker/containerd/daemon/io.containerd.runtime.v2.task/moby/42a1a1ed5912a7227148e997f442e7ab2e5cc3558aa3471548223c5888c9b142 pid=141080
mai 30 15:31:41 curie systemd[1]: run-docker-runtime\x2drunc-moby-42a1a1ed5912a7227148e997f442e7ab2e5cc3558aa3471548223c5888c9b142-runc.ZVcjvl.mount: Succeeded.
mai 30 15:31:41 curie systemd[5287]: run-docker-runtime\x2drunc-moby-42a1a1ed5912a7227148e997f442e7ab2e5cc3558aa3471548223c5888c9b142-runc.ZVcjvl.mount: Succeeded.
mai 30 15:31:41 curie systemd[1]: Started libcontainer container 42a1a1ed5912a7227148e997f442e7ab2e5cc3558aa3471548223c5888c9b142.
mai 30 15:31:45 curie systemd[1]: docker-42a1a1ed5912a7227148e997f442e7ab2e5cc3558aa3471548223c5888c9b142.scope: Succeeded.
mai 30 15:31:45 curie dockerd[3199]: time="2022-05-30T15:31:45.883019128-04:00" level=info msg="shim disconnected" id=42a1a1ed5912a7227148e997f442e7ab2e5cc3558aa3471548223c5888c9b142
mai 30 15:31:45 curie dockerd[1726]: time="2022-05-30T15:31:45.883064491-04:00" level=info msg="ignoring event" container=42a1a1ed5912a7227148e997f442e7ab2e5cc3558aa3471548223c5888c9b142 module=libcontainerd namespace=moby topic=/tasks/delete type="*events.TaskDelete"
mai 30 15:31:45 curie systemd[1]: run-docker-netns-e45f5cf5f465.mount: Succeeded.
mai 30 15:31:45 curie systemd[5287]: run-docker-netns-e45f5cf5f465.mount: Succeeded.
mai 30 15:31:45 curie systemd[1]: var-lib-docker-zfs-graph-41ce08fb7a1d3a9c101694b82722f5621c0b4819bd1d9f070933fd1e00543cdf.mount: Succeeded.
mai 30 15:31:45 curie systemd[5287]: var-lib-docker-zfs-graph-41ce08fb7a1d3a9c101694b82722f5621c0b4819bd1d9f070933fd1e00543cdf.mount: Succeeded.
That's double or triple the run time, from 2 seconds to 6
seconds. Most of the time is spent in run time, inside the
container. Here's the breakdown:
container setup: ~2 seconds
container run: ~4 seconds
container teardown: ~1 second
total run time: about ~6-7 seconds
That's a two- to three-fold increase! Clearly something is going on
here that I should tweak. It's possible that code path is less
optimized in Docker. I also worry about podman, but apparently it
also supports ZFS backends. Possibly it would perform better, but
at this stage I wouldn't have a good comparison: maybe it would have
performed better on non-ZFS as well...
Interactivity
While doing the offsite backups (below), the system became somewhat
"sluggish". I felt everything was slow, and I estimate it introduced
~50ms latency in any input device.
Arguably, those are all USB and the external drive was connected
through USB, but I suspect the ZFS drivers are not as well tuned with
the scheduler as the regular filesystem drivers...
Recovery procedures
For test purposes, I unmounted all systems during the procedure:
umount /mnt/boot/efi /mnt/boot/run
umount -a -t zfs
zpool export -a
And disconnected the drive, to see how I would recover this system
from another Linux system in case of a total motherboard failure.
To import an existing pool, plug the device, then import the pool with
an alternate root, so it doesn't mount over your existing filesystems,
then you mount the root filesystem and all the others:
zpool import -l -a -R /mnt &&
zfs mount rpool/ROOT/debian &&
zfs mount -a &&
mount /dev/sdc2 /mnt/boot/efi &&
mount -t tmpfs tmpfs /mnt/run &&
mkdir /mnt/run/lock
Offsite backup
Part of the goal of using ZFS is to simplify and harden backups. I
wanted to experiment with shorter recovery times specifically both
point in time recovery objective and recovery time objective
and faster incremental backups.
This is, therefore, part of my backup services.
This section documents how an external NVMe enclosure was setup in a
pool to mirror the datasets from my workstation.
The final setup should include syncoid copying datasets to the backup
server regularly, but I haven't finished that configuration yet.
Partitioning
The above partitioning procedure used sgdisk, but I couldn't figure
out how to do this with sgdisk, so this uses sfdisk to dump the
partition from the first disk to an external, identical drive:
First sync
I used syncoid to copy all pools over to the external device. syncoid
is a thing that's part of the sanoid project which is
specifically designed to sync snapshots between pool, typically over
SSH links but it can also operate locally.
The sanoid command had a --readonly argument to simulate changes,
but syncoid didn't so I tried to fix that with an upstream PR.
It seems it would be better to do this by hand, but this was much
easier. The full first sync was:
root@curie:/home/anarcat# ./bin/syncoid -r bpool bpool-tubman
CRITICAL ERROR: Target bpool-tubman exists but has no snapshots matching with bpool!
Replication to target would require destroying existing
target. Cowardly refusing to destroy your existing target.
NOTE: Target bpool-tubman dataset is < 64MB used - did you mistakenly run
zfs create bpool-tubman on the target? ZFS initial
replication must be to a NON EXISTENT DATASET, which will
then be CREATED BY the initial replication process.
INFO: Sending oldest full snapshot bpool/BOOT@test (~ 42 KB) to new target filesystem:
44.2KiB 0:00:00 [4.19MiB/s] [========================================================================================================================] 103%
INFO: Updating new target filesystem with incremental bpool/BOOT@test ... syncoid_curie_2022-05-30:12:50:39 (~ 4 KB):
2.13KiB 0:00:00 [ 114KiB/s] [===============================================================> ] 53%
INFO: Sending oldest full snapshot bpool/BOOT/debian@install (~ 126.0 MB) to new target filesystem:
126MiB 0:00:00 [ 308MiB/s] [=======================================================================================================================>] 100%
INFO: Updating new target filesystem with incremental bpool/BOOT/debian@install ... syncoid_curie_2022-05-30:12:50:39 (~ 113.4 MB):
113MiB 0:00:00 [ 315MiB/s] [=======================================================================================================================>] 100%
root@curie:/home/anarcat# ./bin/syncoid -r rpool rpool-tubman
CRITICAL ERROR: Target rpool-tubman exists but has no snapshots matching with rpool!
Replication to target would require destroying existing
target. Cowardly refusing to destroy your existing target.
NOTE: Target rpool-tubman dataset is < 64MB used - did you mistakenly run
zfs create rpool-tubman on the target? ZFS initial
replication must be to a NON EXISTENT DATASET, which will
then be CREATED BY the initial replication process.
INFO: Sending oldest full snapshot rpool/ROOT@syncoid_curie_2022-05-30:12:50:51 (~ 69 KB) to new target filesystem:
44.2KiB 0:00:00 [2.44MiB/s] [===========================================================================> ] 63%
INFO: Sending oldest full snapshot rpool/ROOT/debian@install (~ 25.9 GB) to new target filesystem:
25.9GiB 0:03:33 [ 124MiB/s] [=======================================================================================================================>] 100%
INFO: Updating new target filesystem with incremental rpool/ROOT/debian@install ... syncoid_curie_2022-05-30:12:50:52 (~ 3.9 GB):
3.92GiB 0:00:33 [ 119MiB/s] [======================================================================================================================> ] 99%
INFO: Sending oldest full snapshot rpool/home@syncoid_curie_2022-05-30:12:55:04 (~ 276.8 GB) to new target filesystem:
277GiB 0:27:13 [ 174MiB/s] [=======================================================================================================================>] 100%
INFO: Sending oldest full snapshot rpool/home/root@syncoid_curie_2022-05-30:13:22:19 (~ 2.2 GB) to new target filesystem:
2.22GiB 0:00:25 [90.2MiB/s] [=======================================================================================================================>] 100%
INFO: Sending oldest full snapshot rpool/var@syncoid_curie_2022-05-30:13:22:47 (~ 5.6 GB) to new target filesystem:
5.56GiB 0:00:32 [ 176MiB/s] [=======================================================================================================================>] 100%
INFO: Sending oldest full snapshot rpool/var/cache@syncoid_curie_2022-05-30:13:23:22 (~ 627.3 MB) to new target filesystem:
627MiB 0:00:03 [ 169MiB/s] [=======================================================================================================================>] 100%
INFO: Sending oldest full snapshot rpool/var/lib@syncoid_curie_2022-05-30:13:23:28 (~ 69 KB) to new target filesystem:
44.2KiB 0:00:00 [1.40MiB/s] [===========================================================================> ] 63%
INFO: Sending oldest full snapshot rpool/var/lib/docker@syncoid_curie_2022-05-30:13:23:28 (~ 442.6 MB) to new target filesystem:
443MiB 0:00:04 [ 103MiB/s] [=======================================================================================================================>] 100%
INFO: Sending oldest full snapshot rpool/var/lib/docker/05c0de7fabbea60500eaa495d0d82038249f6faa63b12914737c4d71520e62c5@266253254 (~ 6.3 MB) to new target filesystem:
6.49MiB 0:00:00 [12.9MiB/s] [========================================================================================================================] 102%
INFO: Updating new target filesystem with incremental rpool/var/lib/docker/05c0de7fabbea60500eaa495d0d82038249f6faa63b12914737c4d71520e62c5@266253254 ... syncoid_curie_2022-05-30:13:23:34 (~ 4 KB):
1.52KiB 0:00:00 [27.6KiB/s] [============================================> ] 38%
INFO: Sending oldest full snapshot rpool/var/lib/flatpak@syncoid_curie_2022-05-30:13:23:36 (~ 2.0 GB) to new target filesystem:
2.00GiB 0:00:17 [ 115MiB/s] [=======================================================================================================================>] 100%
INFO: Sending oldest full snapshot rpool/var/tmp@syncoid_curie_2022-05-30:13:23:55 (~ 57.0 MB) to new target filesystem:
61.8MiB 0:00:01 [45.0MiB/s] [========================================================================================================================] 108%
INFO: Clone is recreated on target rpool-tubman/var/lib/docker/ed71ddd563a779ba6fb37b3b1d0cc2c11eca9b594e77b4b234867ebcb162b205 based on rpool/var/lib/docker/05c0de7fabbea60500eaa495d0d82038249f6faa63b12914737c4d71520e62c5@266253254
INFO: Sending oldest full snapshot rpool/var/lib/docker/ed71ddd563a779ba6fb37b3b1d0cc2c11eca9b594e77b4b234867ebcb162b205@syncoid_curie_2022-05-30:13:23:58 (~ 218.6 MB) to new target filesystem:
219MiB 0:00:01 [ 151MiB/s] [=======================================================================================================================>] 100%
Funny how the CRITICAL ERROR doesn't actually stop syncoid and it
just carries on merrily doing when it's telling you it's "cowardly
refusing to destroy your existing target"... Maybe that's because my pull
request broke something though...
During the transfer, the computer was very sluggish: everything feels
like it has ~30-50ms latency extra:
anarcat@curie:sanoid$ LANG=C top -b -n 1 head -20
top - 13:07:05 up 6 days, 4:01, 1 user, load average: 16.13, 16.55, 11.83
Tasks: 606 total, 6 running, 598 sleeping, 0 stopped, 2 zombie
%Cpu(s): 18.8 us, 72.5 sy, 1.2 ni, 5.0 id, 1.2 wa, 0.0 hi, 1.2 si, 0.0 st
MiB Mem : 15898.4 total, 1387.6 free, 13170.0 used, 1340.8 buff/cache
MiB Swap: 0.0 total, 0.0 free, 0.0 used. 1319.8 avail Mem
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
70 root 20 0 0 0 0 S 83.3 0.0 6:12.67 kswapd0
4024878 root 20 0 282644 96432 10288 S 44.4 0.6 0:11.43 puppet
3896136 root 20 0 35328 16528 48 S 22.2 0.1 2:08.04 mbuffer
3896135 root 20 0 10328 776 168 R 16.7 0.0 1:22.93 zfs
3896138 root 20 0 10588 788 156 R 16.7 0.0 1:49.30 zfs
350 root 0 -20 0 0 0 R 11.1 0.0 1:03.53 z_rd_int
351 root 0 -20 0 0 0 S 11.1 0.0 1:04.15 z_rd_int
3896137 root 20 0 4384 352 244 R 11.1 0.0 0:44.73 pv
4034094 anarcat 30 10 20028 13960 2428 S 11.1 0.1 0:00.70 mbsync
4036539 anarcat 20 0 9604 3464 2408 R 11.1 0.0 0:00.04 top
352 root 0 -20 0 0 0 S 5.6 0.0 1:03.64 z_rd_int
353 root 0 -20 0 0 0 S 5.6 0.0 1:03.64 z_rd_int
354 root 0 -20 0 0 0 S 5.6 0.0 1:04.01 z_rd_int
I wonder how much of that is due to syncoid, particularly because I
often saw mbuffer and pv in there which are not strictly necessary
to do those kind of operations, as far as I understand.
Once that's done, export the pools to disconnect the drive:
Monitoring
ZFS should be monitoring your pools regularly. Normally, the [[!debman
zed]] daemon monitors all ZFS events. It is the thing that will report
when a scrub failed, for example. See this configuration guide.
Scrubs should be regularly scheduled to ensure consistency of the
pool. This can be done in newer zfsutils-linux versions
(bullseye-backports or bookworm) with one of those, depending on the
desired frequency:
When the scrub runs, if it finds anything it will send an event which
will get picked up by the zed daemon which will then send a
notification, see below for an example.
TODO: deploy on curie, if possible (probably not because no RAID)
TODO: this should be in Puppet
Scrub warning example
So what happens when problems are found? Here's an example of how I
dealt with an error I received.
After setting up another server (tubman) with ZFS, I
eventually ended up getting a warning from the ZFS toolchain.
Date: Sun, 09 Oct 2022 00:58:08 -0400
From: root <root@anarc.at>
To: root@anarc.at
Subject: ZFS scrub_finish event for rpool on tubman
ZFS has finished a scrub:
eid: 39536
class: scrub_finish
host: tubman
time: 2022-10-09 00:58:07-0400
pool: rpool
state: ONLINE
status: One or more devices has experienced an unrecoverable error. An
attempt was made to correct the error. Applications are unaffected.
action: Determine if the device needs to be replaced, and clear the errors
using 'zpool clear' or replace the device with 'zpool replace'.
see: https://openzfs.github.io/openzfs-docs/msg/ZFS-8000-9P
scan: scrub repaired 0B in 00:33:57 with 0 errors on Sun Oct 9 00:58:07 2022
config:
NAME STATE READ WRITE CKSUM
rpool ONLINE 0 0 0
mirror-0 ONLINE 0 0 0
sdb4 ONLINE 0 1 0
sdc4 ONLINE 0 0 0
cache
sda3 ONLINE 0 0 0
errors: No known data errors
This, in itself, is a little worrisome. But it helpfully links to this
more detailed documentation (and props up there: the link still
works) which explains this is a "minor" problem (something that could
be included in the report).
In this case, this happened on a server setup on 2021-04-28, but the
disks and server hardware are much older. The server itself
(marcosv1) was built
around 2011, over 10 years ago now. The hard drive in question is:
root@tubman:~# smartctl -i -qnoserial /dev/sdb
smartctl 7.2 2020-12-30 r5155 [x86_64-linux-5.10.0-15-amd64] (local build)
Copyright (C) 2002-20, Bruce Allen, Christian Franke, www.smartmontools.org
=== START OF INFORMATION SECTION ===
Model Family: Seagate BarraCuda 3.5
Device Model: ST4000DM004-2CV104
Firmware Version: 0001
User Capacity: 4,000,787,030,016 bytes [4.00 TB]
Sector Sizes: 512 bytes logical, 4096 bytes physical
Rotation Rate: 5425 rpm
Form Factor: 3.5 inches
Device is: In smartctl database [for details use: -P show]
ATA Version is: ACS-3 T13/2161-D revision 5
SATA Version is: SATA 3.1, 6.0 Gb/s (current: 3.0 Gb/s)
Local Time is: Tue Oct 11 11:02:32 2022 EDT
SMART support is: Available - device has SMART capability.
SMART support is: Enabled
That's over a year of power on, which shouldn't be so bad. It has
written about 10TB of data (21107792664 LBAs * 512 byte/LBA), which
is about two full writes. According to its specification, this
device is supposed to support 55 TB/year of writes, so we're far below
spec. Note that are still far from the "non-recoverable read error per
bits" spec (1 per 10E15), as we've basically read 13E12 bits
(3201579750 LBAs * 512 byte/LBA = 13E12 bits).
It's likely this disk was made in 2018, so it is in its fourth
year.
Interestingly, /dev/sdc is also a Seagate drive, but of a different
series:
root@tubman:~# smartctl -qnoserial -i /dev/sdb
smartctl 7.2 2020-12-30 r5155 [x86_64-linux-5.10.0-15-amd64] (local build)
Copyright (C) 2002-20, Bruce Allen, Christian Franke, www.smartmontools.org
=== START OF INFORMATION SECTION ===
Model Family: Seagate BarraCuda 3.5
Device Model: ST4000DM004-2CV104
Firmware Version: 0001
User Capacity: 4,000,787,030,016 bytes [4.00 TB]
Sector Sizes: 512 bytes logical, 4096 bytes physical
Rotation Rate: 5425 rpm
Form Factor: 3.5 inches
Device is: In smartctl database [for details use: -P show]
ATA Version is: ACS-3 T13/2161-D revision 5
SATA Version is: SATA 3.1, 6.0 Gb/s (current: 3.0 Gb/s)
Local Time is: Tue Oct 11 11:21:35 2022 EDT
SMART support is: Available - device has SMART capability.
SMART support is: Enabled
It has seen much more reads than the other disk which is also interesting:
That's 4 years of Head_Flying_Hours, and over 4 years (4 years and
48 days) of Power_On_Hours. The copyright date on that drive's
specs goes back to 2016, so it's a much older drive.
SMART self-test succeeded.
Remaining issues
TODO: move send/receive backups to offsite host, see also
zfs for alternatives to syncoid/sanoid there
TODO: document this somewhere: bpool and rpool are both pools and
datasets. that's pretty confusing, but also very useful because it
allows for pool-wide recursive snapshots, which are used for the
backup system
fio improvements
I really want to improve my experience with fio. Right now, I'm just
cargo-culting stuff from other folks and I don't really like
it. stressant is a good example of my struggles, in the sense
that it doesn't really work that well for disk tests.
I would love to have just a single .fio job file that lists multiple
jobs to run serially. For example, this file describes the above
workload pretty well:
[global]
# cargo-culting Salter
fallocate=none
ioengine=posixaio
runtime=60
time_based=1
end_fsync=1
stonewall=1
group_reporting=1
# no need to drop caches, done by default
# invalidate=1
# Single 4KiB random read/write process
[randread-4k-4g-1x]
rw=randread
bs=4k
size=4g
numjobs=1
iodepth=1
[randwrite-4k-4g-1x]
rw=randwrite
bs=4k
size=4g
numjobs=1
iodepth=1
# 16 parallel 64KiB random read/write processes:
[randread-64k-256m-16x]
rw=randread
bs=64k
size=256m
numjobs=16
iodepth=16
[randwrite-64k-256m-16x]
rw=randwrite
bs=64k
size=256m
numjobs=16
iodepth=16
# Single 1MiB random read/write process
[randread-1m-16g-1x]
rw=randread
bs=1m
size=16g
numjobs=1
iodepth=1
[randwrite-1m-16g-1x]
rw=randwrite
bs=1m
size=16g
numjobs=1
iodepth=1
... except the jobs are actually started in parallel, even though they
are stonewall'd, as far as I can tell by the reports. I sent a
mail to the fio mailing list for clarification.
It looks like the jobs are started in parallel, but actual
(correctly) run serially. It seems like this might just be a matter of
reporting the right timestamps in the end, although it does feel like
starting all the processes (even if not doing any work yet) could
skew the results.
Hangs during procedure
During the procedure, it happened a few times where any ZFS command
would completely hang. It seems that using an external USB drive to
sync stuff didn't work so well: sometimes it would reconnect under a
different device (from sdc to sdd, for example), and this would
greatly confuse ZFS.
Here, for example, is sdd reappearing out of the blue:
May 19 11:22:53 curie kernel: [ 699.820301] scsi host4: uas
May 19 11:22:53 curie kernel: [ 699.820544] usb 2-1: authorized to connect
May 19 11:22:53 curie kernel: [ 699.922433] scsi 4:0:0:0: Direct-Access ROG ESD-S1C 0 PQ: 0 ANSI: 6
May 19 11:22:53 curie kernel: [ 699.923235] sd 4:0:0:0: Attached scsi generic sg2 type 0
May 19 11:22:53 curie kernel: [ 699.923676] sd 4:0:0:0: [sdd] 1953525168 512-byte logical blocks: (1.00 TB/932 GiB)
May 19 11:22:53 curie kernel: [ 699.923788] sd 4:0:0:0: [sdd] Write Protect is off
May 19 11:22:53 curie kernel: [ 699.923949] sd 4:0:0:0: [sdd] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
May 19 11:22:53 curie kernel: [ 699.924149] sd 4:0:0:0: [sdd] Optimal transfer size 33553920 bytes
May 19 11:22:53 curie kernel: [ 699.961602] sdd: sdd1 sdd2 sdd3 sdd4
May 19 11:22:53 curie kernel: [ 699.996083] sd 4:0:0:0: [sdd] Attached SCSI disk
Next time I run a ZFS command (say zpool list), the command
completely hangs (D state) and this comes up in the logs:
May 19 11:34:21 curie kernel: [ 1387.914843] zio pool=bpool vdev=/dev/sdc3 error=5 type=2 offset=71344128 size=4096 flags=184880
May 19 11:34:21 curie kernel: [ 1387.914859] zio pool=bpool vdev=/dev/sdc3 error=5 type=2 offset=205565952 size=4096 flags=184880
May 19 11:34:21 curie kernel: [ 1387.914874] zio pool=bpool vdev=/dev/sdc3 error=5 type=2 offset=272789504 size=4096 flags=184880
May 19 11:34:21 curie kernel: [ 1387.914906] zio pool=bpool vdev=/dev/sdc3 error=5 type=1 offset=270336 size=8192 flags=b08c1
May 19 11:34:21 curie kernel: [ 1387.914932] zio pool=bpool vdev=/dev/sdc3 error=5 type=1 offset=1073225728 size=8192 flags=b08c1
May 19 11:34:21 curie kernel: [ 1387.914948] zio pool=bpool vdev=/dev/sdc3 error=5 type=1 offset=1073487872 size=8192 flags=b08c1
May 19 11:34:21 curie kernel: [ 1387.915165] zio pool=bpool vdev=/dev/sdc3 error=5 type=2 offset=272793600 size=4096 flags=184880
May 19 11:34:21 curie kernel: [ 1387.915183] zio pool=bpool vdev=/dev/sdc3 error=5 type=2 offset=339853312 size=4096 flags=184880
May 19 11:34:21 curie kernel: [ 1387.915648] WARNING: Pool 'bpool' has encountered an uncorrectable I/O failure and has been suspended.
May 19 11:34:21 curie kernel: [ 1387.915648]
May 19 11:37:25 curie kernel: [ 1571.558614] task:txg_sync state:D stack: 0 pid: 997 ppid: 2 flags:0x00004000
May 19 11:37:25 curie kernel: [ 1571.558623] Call Trace:
May 19 11:37:25 curie kernel: [ 1571.558640] __schedule+0x282/0x870
May 19 11:37:25 curie kernel: [ 1571.558650] schedule+0x46/0xb0
May 19 11:37:25 curie kernel: [ 1571.558670] schedule_timeout+0x8b/0x140
May 19 11:37:25 curie kernel: [ 1571.558675] ? __next_timer_interrupt+0x110/0x110
May 19 11:37:25 curie kernel: [ 1571.558678] io_schedule_timeout+0x4c/0x80
May 19 11:37:25 curie kernel: [ 1571.558689] __cv_timedwait_common+0x12b/0x160 [spl]
May 19 11:37:25 curie kernel: [ 1571.558694] ? add_wait_queue_exclusive+0x70/0x70
May 19 11:37:25 curie kernel: [ 1571.558702] __cv_timedwait_io+0x15/0x20 [spl]
May 19 11:37:25 curie kernel: [ 1571.558816] zio_wait+0x129/0x2b0 [zfs]
May 19 11:37:25 curie kernel: [ 1571.558929] dsl_pool_sync+0x461/0x4f0 [zfs]
May 19 11:37:25 curie kernel: [ 1571.559032] spa_sync+0x575/0xfa0 [zfs]
May 19 11:37:25 curie kernel: [ 1571.559138] ? spa_txg_history_init_io+0x101/0x110 [zfs]
May 19 11:37:25 curie kernel: [ 1571.559245] txg_sync_thread+0x2e0/0x4a0 [zfs]
May 19 11:37:25 curie kernel: [ 1571.559354] ? txg_fini+0x240/0x240 [zfs]
May 19 11:37:25 curie kernel: [ 1571.559366] thread_generic_wrapper+0x6f/0x80 [spl]
May 19 11:37:25 curie kernel: [ 1571.559376] ? __thread_exit+0x20/0x20 [spl]
May 19 11:37:25 curie kernel: [ 1571.559379] kthread+0x11b/0x140
May 19 11:37:25 curie kernel: [ 1571.559382] ? __kthread_bind_mask+0x60/0x60
May 19 11:37:25 curie kernel: [ 1571.559386] ret_from_fork+0x22/0x30
May 19 11:37:25 curie kernel: [ 1571.559401] task:zed state:D stack: 0 pid: 1564 ppid: 1 flags:0x00000000
May 19 11:37:25 curie kernel: [ 1571.559404] Call Trace:
May 19 11:37:25 curie kernel: [ 1571.559409] __schedule+0x282/0x870
May 19 11:37:25 curie kernel: [ 1571.559412] ? __kmalloc_node+0x141/0x2b0
May 19 11:37:25 curie kernel: [ 1571.559417] schedule+0x46/0xb0
May 19 11:37:25 curie kernel: [ 1571.559420] schedule_preempt_disabled+0xa/0x10
May 19 11:37:25 curie kernel: [ 1571.559424] __mutex_lock.constprop.0+0x133/0x460
May 19 11:37:25 curie kernel: [ 1571.559435] ? nvlist_xalloc.part.0+0x68/0xc0 [znvpair]
May 19 11:37:25 curie kernel: [ 1571.559537] spa_all_configs+0x41/0x120 [zfs]
May 19 11:37:25 curie kernel: [ 1571.559644] zfs_ioc_pool_configs+0x17/0x70 [zfs]
May 19 11:37:25 curie kernel: [ 1571.559752] zfsdev_ioctl_common+0x697/0x870 [zfs]
May 19 11:37:25 curie kernel: [ 1571.559758] ? _copy_from_user+0x28/0x60
May 19 11:37:25 curie kernel: [ 1571.559860] zfsdev_ioctl+0x53/0xe0 [zfs]
May 19 11:37:25 curie kernel: [ 1571.559866] __x64_sys_ioctl+0x83/0xb0
May 19 11:37:25 curie kernel: [ 1571.559869] do_syscall_64+0x33/0x80
May 19 11:37:25 curie kernel: [ 1571.559873] entry_SYSCALL_64_after_hwframe+0x44/0xa9
May 19 11:37:25 curie kernel: [ 1571.559876] RIP: 0033:0x7fcf0ef32cc7
May 19 11:37:25 curie kernel: [ 1571.559878] RSP: 002b:00007fcf0e181618 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
May 19 11:37:25 curie kernel: [ 1571.559881] RAX: ffffffffffffffda RBX: 000055b212f972a0 RCX: 00007fcf0ef32cc7
May 19 11:37:25 curie kernel: [ 1571.559883] RDX: 00007fcf0e181640 RSI: 0000000000005a04 RDI: 000000000000000b
May 19 11:37:25 curie kernel: [ 1571.559885] RBP: 00007fcf0e184c30 R08: 00007fcf08016810 R09: 00007fcf08000080
May 19 11:37:25 curie kernel: [ 1571.559886] R10: 0000000000080000 R11: 0000000000000246 R12: 000055b212f972a0
May 19 11:37:25 curie kernel: [ 1571.559888] R13: 0000000000000000 R14: 00007fcf0e181640 R15: 0000000000000000
May 19 11:37:25 curie kernel: [ 1571.559980] task:zpool state:D stack: 0 pid:11815 ppid: 3816 flags:0x00004000
May 19 11:37:25 curie kernel: [ 1571.559983] Call Trace:
May 19 11:37:25 curie kernel: [ 1571.559988] __schedule+0x282/0x870
May 19 11:37:25 curie kernel: [ 1571.559992] schedule+0x46/0xb0
May 19 11:37:25 curie kernel: [ 1571.559995] io_schedule+0x42/0x70
May 19 11:37:25 curie kernel: [ 1571.560004] cv_wait_common+0xac/0x130 [spl]
May 19 11:37:25 curie kernel: [ 1571.560008] ? add_wait_queue_exclusive+0x70/0x70
May 19 11:37:25 curie kernel: [ 1571.560118] txg_wait_synced_impl+0xc9/0x110 [zfs]
May 19 11:37:25 curie kernel: [ 1571.560223] txg_wait_synced+0xc/0x40 [zfs]
May 19 11:37:25 curie kernel: [ 1571.560325] spa_export_common+0x4cd/0x590 [zfs]
May 19 11:37:25 curie kernel: [ 1571.560430] ? zfs_log_history+0x9c/0xf0 [zfs]
May 19 11:37:25 curie kernel: [ 1571.560537] zfsdev_ioctl_common+0x697/0x870 [zfs]
May 19 11:37:25 curie kernel: [ 1571.560543] ? _copy_from_user+0x28/0x60
May 19 11:37:25 curie kernel: [ 1571.560644] zfsdev_ioctl+0x53/0xe0 [zfs]
May 19 11:37:25 curie kernel: [ 1571.560649] __x64_sys_ioctl+0x83/0xb0
May 19 11:37:25 curie kernel: [ 1571.560653] do_syscall_64+0x33/0x80
May 19 11:37:25 curie kernel: [ 1571.560656] entry_SYSCALL_64_after_hwframe+0x44/0xa9
May 19 11:37:25 curie kernel: [ 1571.560659] RIP: 0033:0x7fdc23be2cc7
May 19 11:37:25 curie kernel: [ 1571.560661] RSP: 002b:00007ffc8c792478 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
May 19 11:37:25 curie kernel: [ 1571.560664] RAX: ffffffffffffffda RBX: 000055942ca49e20 RCX: 00007fdc23be2cc7
May 19 11:37:25 curie kernel: [ 1571.560666] RDX: 00007ffc8c792490 RSI: 0000000000005a03 RDI: 0000000000000003
May 19 11:37:25 curie kernel: [ 1571.560667] RBP: 00007ffc8c795e80 R08: 00000000ffffffff R09: 00007ffc8c792310
May 19 11:37:25 curie kernel: [ 1571.560669] R10: 000055942ca49e30 R11: 0000000000000246 R12: 00007ffc8c792490
May 19 11:37:25 curie kernel: [ 1571.560671] R13: 000055942ca49e30 R14: 000055942aed2c20 R15: 00007ffc8c795a40
Here's another example, where you see the USB controller bleeping out
and back into existence:
mai 19 11:38:39 curie kernel: usb 2-1: USB disconnect, device number 2
mai 19 11:38:39 curie kernel: sd 4:0:0:0: [sdd] Synchronizing SCSI cache
mai 19 11:38:39 curie kernel: sd 4:0:0:0: [sdd] Synchronize Cache(10) failed: Result: hostbyte=DID_ERROR driverbyte=DRIVER_OK
mai 19 11:39:25 curie kernel: INFO: task zed:1564 blocked for more than 241 seconds.
mai 19 11:39:25 curie kernel: Tainted: P IOE 5.10.0-14-amd64 #1 Debian 5.10.113-1
mai 19 11:39:25 curie kernel: "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
mai 19 11:39:25 curie kernel: task:zed state:D stack: 0 pid: 1564 ppid: 1 flags:0x00000000
mai 19 11:39:25 curie kernel: Call Trace:
mai 19 11:39:25 curie kernel: __schedule+0x282/0x870
mai 19 11:39:25 curie kernel: ? __kmalloc_node+0x141/0x2b0
mai 19 11:39:25 curie kernel: schedule+0x46/0xb0
mai 19 11:39:25 curie kernel: schedule_preempt_disabled+0xa/0x10
mai 19 11:39:25 curie kernel: __mutex_lock.constprop.0+0x133/0x460
mai 19 11:39:25 curie kernel: ? nvlist_xalloc.part.0+0x68/0xc0 [znvpair]
mai 19 11:39:25 curie kernel: spa_all_configs+0x41/0x120 [zfs]
mai 19 11:39:25 curie kernel: zfs_ioc_pool_configs+0x17/0x70 [zfs]
mai 19 11:39:25 curie kernel: zfsdev_ioctl_common+0x697/0x870 [zfs]
mai 19 11:39:25 curie kernel: ? _copy_from_user+0x28/0x60
mai 19 11:39:25 curie kernel: zfsdev_ioctl+0x53/0xe0 [zfs]
mai 19 11:39:25 curie kernel: __x64_sys_ioctl+0x83/0xb0
mai 19 11:39:25 curie kernel: do_syscall_64+0x33/0x80
mai 19 11:39:25 curie kernel: entry_SYSCALL_64_after_hwframe+0x44/0xa9
mai 19 11:39:25 curie kernel: RIP: 0033:0x7fcf0ef32cc7
mai 19 11:39:25 curie kernel: RSP: 002b:00007fcf0e181618 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
mai 19 11:39:25 curie kernel: RAX: ffffffffffffffda RBX: 000055b212f972a0 RCX: 00007fcf0ef32cc7
mai 19 11:39:25 curie kernel: RDX: 00007fcf0e181640 RSI: 0000000000005a04 RDI: 000000000000000b
mai 19 11:39:25 curie kernel: RBP: 00007fcf0e184c30 R08: 00007fcf08016810 R09: 00007fcf08000080
mai 19 11:39:25 curie kernel: R10: 0000000000080000 R11: 0000000000000246 R12: 000055b212f972a0
mai 19 11:39:25 curie kernel: R13: 0000000000000000 R14: 00007fcf0e181640 R15: 0000000000000000
mai 19 11:39:25 curie kernel: INFO: task zpool:11815 blocked for more than 241 seconds.
mai 19 11:39:25 curie kernel: Tainted: P IOE 5.10.0-14-amd64 #1 Debian 5.10.113-1
mai 19 11:39:25 curie kernel: "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
mai 19 11:39:25 curie kernel: task:zpool state:D stack: 0 pid:11815 ppid: 2621 flags:0x00004004
mai 19 11:39:25 curie kernel: Call Trace:
mai 19 11:39:25 curie kernel: __schedule+0x282/0x870
mai 19 11:39:25 curie kernel: schedule+0x46/0xb0
mai 19 11:39:25 curie kernel: io_schedule+0x42/0x70
mai 19 11:39:25 curie kernel: cv_wait_common+0xac/0x130 [spl]
mai 19 11:39:25 curie kernel: ? add_wait_queue_exclusive+0x70/0x70
mai 19 11:39:25 curie kernel: txg_wait_synced_impl+0xc9/0x110 [zfs]
mai 19 11:39:25 curie kernel: txg_wait_synced+0xc/0x40 [zfs]
mai 19 11:39:25 curie kernel: spa_export_common+0x4cd/0x590 [zfs]
mai 19 11:39:25 curie kernel: ? zfs_log_history+0x9c/0xf0 [zfs]
mai 19 11:39:25 curie kernel: zfsdev_ioctl_common+0x697/0x870 [zfs]
mai 19 11:39:25 curie kernel: ? _copy_from_user+0x28/0x60
mai 19 11:39:25 curie kernel: zfsdev_ioctl+0x53/0xe0 [zfs]
mai 19 11:39:25 curie kernel: __x64_sys_ioctl+0x83/0xb0
mai 19 11:39:25 curie kernel: do_syscall_64+0x33/0x80
mai 19 11:39:25 curie kernel: entry_SYSCALL_64_after_hwframe+0x44/0xa9
mai 19 11:39:25 curie kernel: RIP: 0033:0x7fdc23be2cc7
mai 19 11:39:25 curie kernel: RSP: 002b:00007ffc8c792478 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
mai 19 11:39:25 curie kernel: RAX: ffffffffffffffda RBX: 000055942ca49e20 RCX: 00007fdc23be2cc7
mai 19 11:39:25 curie kernel: RDX: 00007ffc8c792490 RSI: 0000000000005a03 RDI: 0000000000000003
mai 19 11:39:25 curie kernel: RBP: 00007ffc8c795e80 R08: 00000000ffffffff R09: 00007ffc8c792310
mai 19 11:39:25 curie kernel: R10: 000055942ca49e30 R11: 0000000000000246 R12: 00007ffc8c792490
mai 19 11:39:25 curie kernel: R13: 000055942ca49e30 R14: 000055942aed2c20 R15: 00007ffc8c795a40
I understand those are rather extreme conditions: I would fully expect
the pool to stop working if the underlying drives disappear. What
doesn't seem acceptable is that a command would completely hang like
this.
References
See the zfs documentation for more information about ZFS,
and tubman for another installation and migration procedure.
Armadillo is a powerful and expressive C++ template library for linear algebra and scientific computing. It aims towards a good balance between speed and ease of use, has a syntax deliberately close to Matlab, and is useful for algorithm development directly in C++, or quick conversion of research code into production environments. RcppArmadillo integrates this library with the R environment and language and is widely used by (currently) 1027 packages other packages on CRAN, downloaded 26.9 million times (per the partial logs from the cloud mirrors of CRAN), and the CSDA paper (preprint / vignette) by Conrad and myself has been cited 503 times according to Google Scholar.
This release reflect as new upstream bugfix release 11.4.2 made recently by Conrad. To accomodate CRAN and their preference for at most a release per month, we held it back since the 11.4.0 release early October. As we usually do, we generally update once upstream Armadillo releases are made. When we do not immediately release to CRAN (in order to lower the release cadence), we make those interim releases available via GitHub source and the Rcpp drat repo.
This release also brings a rearranged, and as we think, simplified layout of the header files. All existing locations are still supported but we will be starting a (very patient and slow) transition at some point.
The full set of changes (since the last CRAN release 0.11.4.0.1) follows.
Changes in RcppArmadillo version 0.11.4.2.1 (2022-11-08)
Upgraded to Armadillo release 11.4.2 (Ship of Theseus)
more robust handling of corner cases in multi-threaded contexts
Internal header organisation with new sub-directories while providing full compatibility via existing paths (#395#396)
Changes in RcppArmadillo version 0.11.4.1.0 (2022-10-10) (GitHub Only)
Upgraded to Armadillo release 11.4.1 (Ship of Theseus)
fix data race in Cube::slice()
workarounds for false warnings emitted by GCC 12 when compiling with FORTIFY_SOURCE enabled (already in RcppArmadillo 0.11.4.0.1 too)
Computer terminals have traditionally had an actual bell
that would ring when a certain control character (the bell
character, typically control-g or \a in an C escape
sequence) would come in the input stream.
That feature actually predates computers altogether, and was present
in Baudot code, "an early character encoding for telegraphy
invented by mile Baudot in the 1870s", itself superseding Morse
code.
Modern terminal emulators have, of course, kept that feature: if
you run this command in a terminal right now:
printf '\a'
... you may hear some annoying beep. Or not. It actually depends on
a lot of factors, including which terminal emulator you're using, how
it's configured, whether you have headphones on, or speakers
connected, or, if you're really old school, a PC speaker, even.
Typically, I have this theory that it does the exact opposite of what
you want, regardless of whether you have configured it or not. That
is, if you want it to make noises, it won't, and if you want it to
stay silent, it will make brutal, annoying noises in moments you would
the least expect. I suspect this is a law in computer science, but I'm
too lazy to come up with a formal definition.
Yet something can be done with this.
Making the bell useful and silent
Many terminal emulators have this feature where they can silence the
bell somehow. It can be turned into a "visual bell" which basically
flashes the screen when a bell arrives. Or that can also be disabled
and the bell is just completely ignored.
What I did instead is turn the bell into a "urgency hint" (part
of the ICCCM standard. In xterm, this is done with this X
resource entry (typically in ~/.Xresources):
XTerm*bellIsUrgent: true
XTerm*visualBell: false
Interestingly, this doesn't clearly say "bell is muted", but it's
effectively what it does. Or maybe it works because I have muted
"System Sounds" in Pulseaudio. Who knows. I do have this in my startup
scripts though:
If the dash or 'off' are given, the bell will be turned off.
Interestingly, you have the option of setting the bell "volume",
"pitch, in hertz, and [...] duration in milliseconds. Note that not
all hardware can vary the bell characteristics." In any case, I think
that's the magic trick to turn the darn thing off.
Now this should send urgency hints to your window manager:
sleep 3 ; printf '\a'
Try it: run the command, switch to another desktop, then wait 3
seconds. You should see the previous desktop show up in red or
something.
In the i3 window manager I am currently using, this is
the default, although I did set the colors (client.urgent and
urgent_workspace in bar.colors).
Other window managers or desktop environments may require different
configurations.
Sending a bell...
Now that, on itself, will only be useful when something sets a
bell. One place I had found a trick like this, long ago, is this
post (dead link, archived) which has various instructions for
different tools. I'll recopy some of them here since the original site
is dead, but credit goes to the Netbuz blog.
(Note that the blog post also features an Awesome WM configuration for
urgency hints.)
Mutt
set beep=yes
set beep_new=yes
Irssi
/set beep_when_window_active ON
/set beep_when_away ON
/set beep_msg_level MSGS DCC DCCMSGS HILIGHT
It was also recommending this setting, but it appears to be deprecated
and gives a warning in modern irssi versions:
/set bell_beeps ON
GNU Screen
This is an important piece of the puzzle, because by default, terminal
multiplexers have their own opinion of what to do with the bell as
well:
# disabled: we want to propagate bell to clients, which should handle
# it in their own terminal settings. this vbell off is also the
# upstream and tmux's default
#
# see also: http://netbuz.org/blog/2011/11/x-bells-and-urgency-hints/
vbell off
# propagate bell from other windows up to the terminal emulator as well
bell_msg 'Bell in window %n^G'
The bell_msg bit is an extra from me: it uses the bell message that
pops up when screen detects a bell in another window to resend the
bell control character up to the running terminal.
This makes it so a bell in any multiplexed window will also
propagate to the parent terminal, which is not the default.
Tmux
Untested, but that is apparently how you do it:
# listen to alerts from all windows
set -g bell-action any
# notice bell in windows
set -g monitor-bell on
# only propagate bell, don't warn user, as it hangs tmux for a second
set -g visual-bell off
# send bell *and* notify when activity (if monitor-activity)
set -g visual-activity both
Note that this config goes beyond what we have in GNU screen in that
inactivity or activity will trigger a bell as well. This might be
useful for cases where you don't have the prompt hack (below) but it
could also very well be very noisy. It will only generate noise when
monitor-activity is enabled though.
bash and shell prompts
Now the icing on cake is to actually send the bell when a command
completes. This is what I use this for the most, actually.
I was previously using undistract-me for this, actually. That was
nice: it would send me a nice desktop notification when a command was
running more than a certain amount of time, and if the window was
unfocused. But configuring this was a nightmare: because it uses a
complex PROMPT_COMMAND in bash, it would conflict with my (already
existing and way too) complex bash prompt, and lead to odd
behaviors. It would also not work for remote commands, of course, as
it wouldn't have access to my local D-BUS to send notifications
(thankfully!).
So instead, what I do now is systematically print a bell whenever a
command terminates, in all my shells. I have this in my
/root/.bashrc on all my servers, deployed in Puppet:
PROMPT_COMMAND='printf "\a"'
Or you can just put it directly in the shell prompt, with something
like:
PS1='\[\a\]'"$PS1"
(I have the equivalent in my own .bashrc, although that thing is
much more complex, featuring multi-command pipeline exit status,
colors, terminal title setting, and more, which should probably
warrant its own blog post.)
This sounds a little bonkers and really noisy, but remember that I
turned off the audible bell. And urgency hints are going to show up
only if the window is unfocused. So it's actually really nice and not
really distracting.
Or, to reuse the undistract-me concept: it allows me to not lose focus
too much when I'm waiting for a long process to complete.
That idea actually came from ahf, so kudos to him on that nice
hack.
Caveats
That is not setup on all the machines I administer for work, that
said. I'm afraid that would be too disruptive for people who do not
have that configuration. This implies that I don't get notifications
for commands that run for a long time on remote servers, most of the
time. That said, I could simply run a command with a trailing printf
'\a' to get a notification.
This might not work in Wayland, your window manager, your desktop
environment, your Linux console, or your telegraphy session.
Marie is a royal bastardess, a product of rape no less, and entirely out
of place in the court in Westminster, where she landed after being kicked
off her mother's farm. She had run the farm since her mother's untimely
death, but there was no way that her relatives would let her inherit. In
court, Marie is too tall, too ugly, and too strange, raised by women who
were said to carry the blood of the fairy M lusine. Eleanor of
Aquitaine's solution to her unwanted house guest is a Papal commission.
Marie is to become the prioress of an abbey.
I am occasionally unpleasantly reminded of why I don't read very much
literary fiction. It's immensely frustrating to read a book in which the
author cares about entirely different things than the reader, and where
the story beats all land wrong.
This is literary historical fiction set in the 12th century. Marie is
Marie de France,
author of the lais about courtly love that are famous primarily due to
their position as early sources for the legends of King Arthur. The lais
are written on-screen very early in this book, but they disappear without
any meaningful impact on the story. Matrix is, instead, about
Shaftesbury Abbey and what it becomes during Marie's time as prioress and
then abbess, following the theory that Marie de France was
Mary of
Shaftesbury.
What I thought I was getting in this book, from numerous reviews and
recommendations, was a story of unexpected competence: how a wild,
unwanted child of seventeen lands at a dilapidated and starving abbey,
entirely against her will, and then over the next sixty years transforms
it into one of the richest abbeys in England. This does happen in this
book, but Groff doesn't seem to care about the details of that
transformation at all.
Instead, Matrix takes the mimetic fiction approach of detailed and
precise description of a few characters, with all of their flaws and
complexities, and with all of the narrative's attention turned to how they
are feeling and what they are seeing. It is also deeply, fully committed
to a Great Man (or in this case a Great Woman) view of history.
Marie is singular. The narrative follows her alone, she makes all the
significant decisions, and the development of the abbey is determined by
her apparently mystical visions. (In typical mimetic fashion, these are
presented as real to her, and the novel takes no position on whether that
reality is objective.) She builds spy networks, maneuvers through local
and church politics, and runs the abbey like her personal kingdom. The
tiny amount of this that is necessarily done by other people is attributed
to Marie's ability to judge character. Other people's motives are simply
steamrolled over and have no effect.
Maddeningly, essentially all of this happens off-screen, and Groff is
completely uninterested in the details of how any of it is accomplished.
Marie decides to do something, the narrative skips forward a year, and it
has happened. She decides to build something, and then it's built. She
decides to collect the rents she's due, the novel gestures vaguely at how
she's intimidating, and then everyone is happily paying up. She builds
spy networks; who cares how? She maneuvers through crises of local and
church politics that are vaguely alluded to, through techniques that are
apparently too uninteresting to bother the reader with.
Instead, the narrative focuses on two things: her deeply dysfunctional,
parasocial relationship with Eleanor, and her tyrannical relationship with
the other nuns. I suspect that Groff would strongly disagree with my
characterization of both of those narratives, and that's the other half of
my problem with this book.
Marie is obsessed with and in love with Eleanor, a completely impossible
love to even talk about, and therefore turns to courtly love from afar as
a model into which she can fit her feelings. While this is the setup for
a tragedy, it's a good idea for a story. But what undermined it for me is
that Marie's obsession seems to be largely physical (she constantly dwells
on Eleanor's beauty), and Eleanor is absolutely horrible to her in every
way: condescending, contemptuous, dismissive, and completely uninterested.
This does change a bit over the course of the book, but not enough to
justify the crush that Marie maintains for this awful person through her
entire life.
And Eleanor is the only person in the book who Marie treats like an equal.
Everyone else is a subordinate, a daughter, a charge, a servant, or a
worker. The nuns of the abbey prosper under her rule, so Marie has ample
reason to justify this to herself, but no one else's opinions or beliefs
matter to her in any significant way. The closest anyone can come to
friendship is to be reliably obedient, perhaps after some initial
objections that Marie overrules. Despite some quite good characterization
of the other nuns, none of the other characters get to do anything.
There is no delight in teamwork, sense of healthy community, or
collaborative problem-solving. It's just all Marie, all the time,
imposing her vision on the world both living and non-living through sheer
force of will.
This just isn't entertaining, at least for me. The writing might be
beautiful, the descriptions detailed and effective, and the research
clearly extensive, but I read books primarily for characters, I read
characters primarily for their relationships, and these relationships are
deeply, horribly unhealthy. They are not, to be clear, unrealistic
(although I do think there's way too much chosen one in Marie and way too
many problems that disappear off-camera); there are certainly people in
the world with dysfunctional obsessive relationships, and there are
charismatic people who overwhelm everyone around them. This is just not
what I want to read about.
You might think, with all I've said above, that I'm spoiling a lot of the
book, but weirdly I don't think I am. Every pattern I mention above is
well-established early in the novel. About the only thing that I'm
spoiling is the hope that any of it is somehow going to change, a hope
that I clung to for rather too long.
This is a great setup for a book, and I wish it were written by a fantasy
author instead of a literary author. Perhaps I'm being too harsh on
literary fiction here, but I feel like fantasy authors are more likely to
write for readers who want to see the growth sequence. If someone is
going to change the world, I want to see how they changed the
world. The mode of fantasy writing tends to think that what people do
(and how they do it) is as interesting or more interesting than what they
feel or what they perceive.
If this idea, with the exact same level of (minor) mysticism and historic
realism, without added magic, were written by, say, Guy Gavriel Kay or
Nicola Griffith, it would be a far different and, in my opinion, a much
better book. In fact, Hildis
part of this book written by Nicola Griffith, and it is a much
better book.
I have seen enough people rave about this book to know that this is a
personal reaction that is not going to be shared by everyone, or possibly
even most people. My theory is that this is due to the different reading
protocols between literary fiction readers and fantasy readers. I put
myself in the latter camp; if you prefer literary fiction, you may like
this much better (and also I'm not sure you'll find my book reviews
useful). I may be wrong, though; maybe there are fantasy readers who
would like this. I will say that the sense of place is very strong and
the writing has all the expected literary strengths of descriptiveness and
rhythm.
But, sadly, this was not at all my thing, and I'm irritated that I wasted
time on it.
Rating: 4 out of 10
A new release 0.4.20 of RProtoBuf arrived on CRAN earlier today. RProtoBuf provides R with bindings for the Google Protocol Buffers ( ProtoBuf ) data encoding and serialization library used and released by Google, and deployed very widely in numerous projects as a language and operating-system agnostic protocol.
This release is somewhat mechanical and in the spirit of many other recent releases. clang-15 is, as more recent compilers do, more stringent on definitions and wants to see some void in argument-less signatures. Happy to oblige. At the same time, GitHub Actions started to nag us about minimum versions of node code so an upgrade to a newer action is warranted, again as with many other affected packages. We also found another http:// URL hiding somewhere so that was cleaned. Lastly, it appears Protocol Buffers themselves moved on and now need / prefer C++17 so were happy to oblige.
The following section from the NEWS.Rd file has full details.
Changes in RProtoBuf version 0.4.20 (2022-11-02)
Two function prototype were updated for clang-15.
GitHub Actions were updated to checkout version 3.
One more http:// URL was updated to https://
The package compilation is now donw with C++17 as the Protocol Buffer headers require it (and clang-15 nags otherwise).
Carpe Jugulum is the 23rd Discworld novel and the 6th witches
novel. I would not recommend reading it before Maskerade, which introduces Agnes.
There are some spoilers for Wyrd
Sisters, Lords and Ladies, and
Maskerade in the setup here and hence in the plot description
below. I don't think they matter that much, but if you're avoiding all
spoilers for earlier books, you may want to skip over this one. (You're
unlikely to want to read it before those books anyway.)
It is time to name the child of the king of Lancre, and in a gesture of
good will and modernization, he has invited his neighbors in Uberwald to
attend. Given that those neighbors are vampires, an open invitation was
perhaps not the wisest choice.
Meanwhile, Granny Weatherwax's invitation has gone missing. On the plus
side, that meant she was home to be summoned to the bedside of a pregnant
woman who was kicked by a cow, where she makes the type of hard decision
that Granny has been making throughout the series. On the minus side, the
apparent snub seems to send her into a spiral of anger at the lack of
appreciation.
Points off right from the start for a plot based on a misunderstanding and
a subsequent refusal of people to simply talk to each other. It is partly
engineered, but still, it's a cheap and irritating plot.
This is an odd book.
The vampires (or vampyres, as the Count wants to use) think of themselves
as modern and sophisticated, making a break from the past by attempting to
overcome such traditional problems as burning up in the sunlight and fear
of religious symbols and garlic. The Count has put his family through
rigorous training and desensitization, deciding such traditional
vulnerabilities are outdated things of the past. He has, however, kept
the belief that vampires are at the top of a natural chain of being,
humans are essentially cattle, and vampires naturally should rule and feed
on the population. Lancre is an attractive new food source. Vampires
also have mind control powers, control the weather, and can put their
minds into magpies.
They are, in short, enemies designed for Granny Weatherwax, the witch
expert in headology. A shame that Granny is apparently off sulking.
Nanny and Agnes may have to handle the vampires on their own, with the
help of Magrat.
One of the things that makes this book odd is that it seemed like
Pratchett was setting up some character growth, giving Agnes a chance to
shine, and giving Nanny Ogg a challenge that she didn't want. This sort
of happens, but then nothing much comes of it. Most of the book is the
vampires preening about how powerful they are and easily conquering
Lancre, while everyone else flails ineffectively. Pratchett does pull
together an ending with some nice set pieces, but that ending doesn't
deliver on any of the changes or developments it felt like the story was
setting up.
We do get a lot of Granny, along with an amusingly earnest priest of Om
(lots of references to Small Gods here,
while firmly establishing it as long-ago history). Granny is one of my
favorite Discworld characters, so I don't mind that, but we've seen Granny
solve a lot of problems before. I wanted to see more of Agnes, who is the
interesting new character and whose dynamic with her inner voice feels
like it has a great deal of unrealized potential.
There is a sharp and condensed version of comparative religion from
Granny, which is probably the strongest part of the book and includes one
of those Discworld quotes that has been widely repeated out of context:
"And sin, young man, is when you treat people as things. Including
yourself. That's what sin is."
"It's a lot more complicated than that "
"No. It ain't. When people say things are a lot more complicated
than that, they means they're getting worried that they won t like the
truth. People as things, that's where it starts."
This loses a bit in context because this book is literally about treating
people as things, and thus the observation feels more obvious when it
arrives in this book than when you encounter it on its own, but it's still
a great quote.
Sadly, I found a lot of this book annoying. One of those annoyances is a
pet peeve that others may or may not share: I have very little patience
for dialogue in phonetically-spelled dialect, and there are two
substantial cases of that here. One is a servant named Igor who speaks
with an affected lisp represented by replacing every ess sound with th,
resulting in lots of this:
"No, my Uncle Igor thtill workth for him. Been thtruck by lightning
three hundred timeth and thtill putth in a full night'th work."
I like Igor as a character (he's essentially a refugee from The
Addams Family, which adds a good counterpoint to the malicious and
arrogant evil of the vampires), but my brain stumbles over words like
"thtill" every time. It's not that I can't decipher it; it's that the
deciphering breaks the flow of reading in a way that I found not at all
fun. It bugged me enough that I started skipping his lines if I couldn't
work them out right away.
The other example is the Nac Mac Feegles, who are... well, in the book,
they're Pictsies and a type of fairy, but they're Scottish Smurfs, right
down to only having one female (at least in this book). They're
entertainingly homicidal, but they all talk like this:
"Ach, hins tak yar scaggie, yer dank yowl callyake!"
I'm from the US and bad with accents and even worse with accents
reproduced in weird spellings, and I'm afraid that I found 95% of
everything said by Nac Mac Feegles completely incomprehensible to the
point where I gave up even trying to read it. (I'm now rather worried
about the Tiffany Aching books and am hoping Pratchett toned the dialect
down a lot, because I'm not sure I can deal with more of this.)
But even apart from the dialect, I thought something was off about the
plot structure of this book. There's a lot of focus on characters who
don't seem to contribute much to the plot resolution. I wanted more of
the varied strengths of Lancre coming together, rather than the focus on
Granny. And the vampires are absurdly powerful, unflappable, smarmy, and
contemptuous of everyone, which makes for threatening villains but also
means spending a lot of narrative time with a Discworld version of
Jacob Rees-Mogg. I
feel like there's enough of that in the news already.
Also, while I will avoid saying too much about the plot, I get very
suspicious when older forms of oppression are presented as good
alternatives to modernizing, rationalist spins on exploitation. I see
what Pratchett was trying to do, and there is an interesting point here
about everyone having personal relationships and knowing their roles (a
long-standing theme of the Lancre Discworld stories). But I think the
reason why there is some nostalgia for older autocracy is that we only
hear about it from stories, and the process of storytelling often creates
emotional distance and a patina of adventure and happy outcomes. Maybe
you can make an argument that classic British imperialism is superior to
smug neoliberalism, but both of them are quite bad and I don't want either
of them.
On a similar note, Nanny Ogg's tyranny over her entire extended clan
continues to be played for laughs, but it's rather unappealing and seems
more abusive the more one thinks about it. I realize the witches are not
intended to be wholly good or uncomplicated moral figures, but I want to
like Nanny, and Pratchett seems to be writing her as likable, even though
she has an astonishing lack of respect for all the people she's related
to. One might even say that she treats them like things.
There are some great bits in this book, and I suspect there are many
people who liked it more than I did. I wouldn't be surprised if it was
someone's favorite Discworld novel. But there were enough bits that
didn't work for me that I thought it averaged out to a middle-of-the-road
entry.
Followed by The Fifth Elephant in publication order. This is the
last regular witches novel, but some of the thematic thread is picked up
by The Wee Free Men, the first Tiffany Aching novel.
Rating: 7 out of 10
A new version 0.1.4 of the tint package arrived at CRAN today. tint provides a style not unlike Tufte for use in html and pdf documents created from markdown. The github repo shows several examples in its README, more as usual in the package documentation.
This is the first release in over two years and it brings a few incremental changes which are described in detail below. The trigger was an email from CRAN desiring compacted pdf files inside the package sources. Which, coincidentally, was made a default about a year ago in the build.r script from littler I use to make packages. But as we had not updated tint this has not reached CRAN. So there.
The set of changes follows, which includes a helpful contributed PR for which we are always thankful. I also improved table rendering, most other changes are packaging mechanics.
Changes in tint version 0.1.4 (2022-11-01)
Added continuous integration at GitHub using run.sh
Non-margin reference in html mode now have an improved display (David Selby and Dirk in #47)
The LaTeX package array is now installed for improved table support (Dirk addressing #48)
Minor updates made the package directory and README.md
GitHub Actions were updated to checkout version 3.
An Bord Plean la conceded that the N6 Galway City Ring Road was in contravention with the National Climate Action plan.
In planning for how we adapt to and mitigate climate change, the UN IPCC AR6 report summarised the state of research globally and modelled nearly 3000 scenarios. The scenarios that kept us below 2 degrees of warming (our Paris agreement commitments) overwhelmingly depended on a move away from car-dependent societies and relied on car-free cities. This report was signed off by every world government.
All EU and national modelling agrees. Accordingly our national Climate Action plan prepares for this global move from cars. It involves a reduction in car traffic, heavy investment in public transport and active travel (walking and cycling). This is essential as the research shows this works, and takes up to 84% less energy for transport in cities.
The N6GCRR plan ignored this, and assumed an increase in car traffic of up to 49%. It in practice would have locked in car dependency when we don t expect there to be enough materials (lithium, etc for batteries) or energy to support that number of cars.
Across Europe and the world we have seen how to move from cars in cities: rail, reliable bus services with dedicated bus corridors; safe, segregated cycle lanes, public space design that enables walking, cycling by keeping distances short the 15 minute city concept of having schools, shops, transport hubs less than 15 minutes from your door. From Paris, removing 40% of cars from the city, now removing a lane from the Peripherique (its ring road), to Korea where expressways in cities have been removed to improve traffic as part of its better design. In Ireland, Dublin has seen a 49% drop in peak-time traffic over recent decades with improved public transport and cycling.
But in Galway we ve ignored all this: not a kilometre of segregated cycle lane added, as there has been an over 80% drop in pupils cycling to school since the 1980s, amplifying traffic as parents commute to drop them to school, making roads more dangerous. No bus route passes over the Quincentennial bridge.
Rather than fix the traffic in Galway, a ring road would induce more traffic as the existing Bothar na dTreabh / Quincintennial bridge did; the current road opened up Knocknacarra for development; a new motorway to Barna would open up Barna for more development.
This is zombie planning: a roads plan from the 1990s lives on, while our County Development Plan has moved the focus to creating new satellite towns in Ardaun and Garraun instead. When all new business wants to be on the East, the new developments need to be near them; new schools are being built on the East, not the West. Rail is being double-tracked to Oranmore (and on to Athenry) to provide commuter rail to Renmore and the growing towns. This reduces the demand for cross-city commuting.
Active travel and public transport development in the city has been stalled awaiting the ring road. This needs to be implemented as quickly as possible. The bus corridor from Salthill (via the schools) to Oranmore (via the businesses at Ballybrit, Parkmore) needs to be completed; this will involve a bus lane on the bridge. The N6GCRR said this was impossible without a new bridge for cars; Dublin s experience with the quays shows otherwise. If we get a fraction of the drop Dublin has seen, the traffic problem will be solved, as shown when the schools are out for holidays.
The NTA and Galway councils have been stuck in thinking from 30 years ago. They assume that traffic cannot be reduced, only added; the ABP inspector agreed, despite overwhelming evidence elsewhere. We need to implement what has been shown to work elsewhere as quickly as possible; the funding is in place but the will and vision to do so has been lacking.
In my home lab(s), I have a handful of machines spread around a few
points of presence, with mostly residential/commercial cable/DSL
uplinks, which means, generally, NAT. This makes monitoring those
devices kind of impossible. While I do punch holes for SSH, using jump
hosts gets old quick, so I'm considering adding a virtual private
network (a "VPN", not a VPN service) so that all machines can
be reachable from everywhere.
I see three ways this can work:
a home-made Wireguard VPN, deployed with Puppet
a Wireguard VPN overlay, with Tailscale or equivalent
IPv6, native or with tunnels
So which one will it be?
Wireguard Puppet modules
As is (unfortunately) typical with Puppet, I found multiple different
modules to talk with Wireguard.
The voxpupuli module seems to be the most promising. The
abaranov module is more popular and has more contributors, but
it has more open issues and PRs.
More critically, the voxpupuli module was written after the abaranov
author didn't respond to a PR from the voxpupuli author trying to
add more automation (namely private key management).
It looks like setting up a wireguard network would be as simple as
this on node A:
This configuration come from this pull request I sent to the
module to document how to use that fact.
Note that the addresses used here are examples that shouldn't be
reused and do not confirm to RFC5737 ("IPv4 Address Blocks
Reserved for Documentation", 192.0.2.0/24 (TEST-NET-1),
198.51.100.0/24 (TEST-NET-2), and 203.0.113.0/24 (TEST-NET-3)) or
RFC3849 ("IPv6 Address Prefix Reserved for Documentation",
2001:DB8::/32), but that's another story.
(To avoid boostrapping problems, the resubmit-facts configuration
could be used so that other nodes facts are more immediately
available.)
One problem with the above approach is that you explicitly need to
take care of routing, network topology, and addressing. This can get
complicated quickly, especially if you have lots of devices, behind
NAT, in multiple locations (which is basically my life at home,
unfortunately).
Concretely, basic Wireguard only support one peer behind
NAT. There are some workarounds for this, but they generally imply
a relay server of some sort, or some custom registry, it's
kind of a mess. And this is where overlay networks like Tailscale come
in.
Tailscale
Tailscale is basically designed to deal with this problem. It's
not fully opensource, but pretty close, and they have an interesting
philosophy behind that. The client is opensource, and there
is an opensource version of the server side, called
headscale. They have recently (late 2022) hired the main headscale
developer while promising to keep supporting it, which is pretty
amazing.
Tailscale provides an overlay network based on Wireguard, where each
peer basically has a peer-to-peer encrypted connexion, with automatic
key rotation. They also ship a multitude of applications and features
on top of that like file sharing, keyless SSH access, and so on. The
authentication layer is based on an existing SSO provider, you
don't just register with Tailscale with new account, you login with
Google, Microsoft, or GitHub (which, really, is still Microsoft).
The Headscale server ships with many features out of that:
Neither project (client or server) is in Debian (RFP 972439 for
the client, none filed yet for the server), which makes deploying this
for my use case rather problematic. Their install instructions
are basically a curl bash but they also provide packages for
various platforms. Their Debian install instructions are
surprisingly good, and check most of the third party checklist
we're trying to establish. (It's missing a pin.)
There's also a Puppet module for tailscale, naturally.
What I find a little disturbing with Tailscale is that you not only
need to trust Tailscale with authorizing your devices, you also
basically delegate that trust also to the SSO provider. So, in my
case, GitHub (or anyone who compromises my account there) can
penetrate the VPN. A little scary.
Tailscale is also kind of an "all or nothing" thing. They have
MagicDNS, file transfers, all sorts of things, but those things
require you to hook up your resolver with Tailscale. In fact,
Tailscale kind of assumes you will use their nameservers, and have
suffered great lengths to figure out how to do that. And
naturally, here, it doesn't seem to work reliably; my resolv.conf
somehow gets replaced and the magic resolution of the ts.net domain
fails.
(I wonder why we can't opt in to just publicly resolve the ts.net
domain. I don't care if someone can enumerate the private IP addreses
or machines in use in my VPN, at least I don't care as much as
fighting with resolv.conf everywhere.)
Because I mostly have access to the routers on the networks I'm on, I
don't think I'll be using tailscale in the long term. But it's pretty
impressive stuff: in the time it took me to even review the Puppet
modules to configure Wireguard (which is what I'll probably end up
doing), I was up and running with Tailscale (but with a broken DNS,
naturally).
(And yes, basic Wireguard won't bring me DNS either, but at least I
won't have to trust Tailscale's Debian packages, and Tailscale, and
Microsoft, and GitHub with this thing.)
IPv6
IPv6 is actually what is supposed to solve this. Not NAT port
forwarding crap, just real IPs everywhere.
The problem is: even though IPv6 adoption is still growing, it's
kind of reaching a plateau at around 40% world-wide, with Canada
lagging behind at 34%. It doesn't help that major ISPs in Canada
(e.g. Bell Canada, Videotron) don't care at all about IPv6
(e.g. Videotron in betasince 2011). So we can't rely on
those companies to do the right thing here.
The typical solution here is often to use a tunnel like HE's
tunnelbroker.net. It's kind of tricky to configure, but once
it's done, it works. You get end-to-end connectivity as long as
everyone on the network is on IPv6.
And that's really where the problem lies here; the second one of
your nodes can't setup such a tunnel, you're kind of stuck and that
tool completely breaks down. IPv6 tunnels also don't give you the kind
of security a VPN provides as well, naturally.
The other downside of a tunnel is you don't really get peer-to-peer
connectivity: you go through the tunnel. So you can expect higher
latencies and possibly lower bandwidth as well. Also, HE.net doesn't
currently charge for this service (and they've been doing this for a
long time), but this could change in the future (just like
Tailscale, that said).
Concretely, the latency difference is rather minimal, Google:
--- ipv6.l.google.com ping statistics ---
10 packets transmitted, 10 received, 0,00% packet loss, time 136,8ms
RTT[ms]: min = 13, median = 14, p(90) = 14, max = 15
--- google.com ping statistics ---
10 packets transmitted, 10 received, 0,00% packet loss, time 136,0ms
RTT[ms]: min = 13, median = 13, p(90) = 14, max = 14
In the case of GitHub, latency is actually lower, interestingly:
--- ipv6.github.com ping statistics ---
10 packets transmitted, 10 received, 0,00% packet loss, time 134,6ms
RTT[ms]: min = 13, median = 13, p(90) = 14, max = 14
--- github.com ping statistics ---
10 packets transmitted, 10 received, 0,00% packet loss, time 293,1ms
RTT[ms]: min = 29, median = 29, p(90) = 29, max = 30
That is because HE.net peers directly with my ISP and Fastly (which
is behind GitHub.com's IPv6, apparently?), so it's only 6 hops
away. While over IPv4, the ping goes over New York, before landing
AWS's Ashburn, Virginia datacenters, for a whopping 13 hops...
I managed setup a HE.net tunnel at home, because I also need IPv6
for other reasons (namely debugging at work). My first attempt at
setting this up in the office failed, but now that I found the
openwrt.org guide, it worked... for a while, and I was able to
produce the above, encouraging, mini benchmarks.
Unfortunately, a few minutes later, IPv6 just went down again. And the
problem with that is that many programs (and especially
OpenSSH) do not respect the Happy Eyeballs protocol (RFC
8305), which means various mysterious "hangs" at random times on
random applications. It's kind of a terrible user experience, on top
of breaking the one thing it's supposed to do, of course, which is to
give me transparent access to all the nodes I maintain.
Even worse, it would still be a problem for other remote nodes I might
setup where I might not have acess to the router to setup the
tunnel. It's also not absolutely clear what happens if you setup the
same tunnel in two places... Presumably, something is smart enough to
distribute only a part of the /48 block selectively, but I don't
really feel like going that far, considering how flaky the setup is
already.
Other options
If this post sounds a little biased towards IPv6 and Wireguard, it's
because it is. I would like everyone to migrate to IPv6 already, and
Wireguard seems like a simple and sound system.
I'm aware of many other options to make VPNs. So before anyone jumps
in and says "but what about...", do know that I have personnally
experimented with:
ipsec, specifically strongswan: hard to configure
(especially configure correctly!), harder even to debug, otherwise
really nice because transparent (e.g. no need for special subnets),
used at work, but also considering a replacement there because it's
a major barrier to entry to train new staff
OpenVPN: mostly used as a client for [VPN service][]s like
Riseup VPN or Mullvad, mostly relevant for client-server
configurations, not really peer-to-peer, shared secrets or TLS,
kind of an hassle to maintain, see also SoftEther for an
alternative implementation
All of those solutions have significant problems and I do not wish to
use any of those for this project.
Also note that Tailscale is only one of many projects laid over
Wireguard to do that kind of thing, see this LWN review for
others (basically NetbBird, Firezone, and Netmaker).
Future work
Those are options that came up after writing this post, and might
warrant further examination in the future.
Meshbird, a "distributed private networking" with little
information about how it actually works other than "encrypted with
strong AES-256"
Nebula, "A scalable overlay networking tool with a focus on
performance, simplicity and security", written by Slack people to
replace IPsec, docs, runs as an overlay for Slack's 50k
node network, only packaged in Debian experimental, lagging behind
upstream (1.4.0, from May 2021 vs upstream's 1.6.1 from September
2022), requires a central CA, Golang, I'm in "wait and see" mode
for now
ouroboros: "peer-to-peer packet network prototype", sounds and
seems complicated
QuickTUN is interesting because it's just a small wrapper
around NaCL, and it's in Debian... but maybe too obscure for my own
good
unetd: Wireguard-based full mesh networking from OpenWRT, not
in Debian
vpncloud: "high performance peer-to-peer mesh VPN over UDP
supporting strong encryption, NAT traversal and a simple
configuration", sounds interesting, not in Debian
Conclusion
Right now, I'm going to deploy Wireguard tunnels with Puppet. It seems
like kind of a pain in the back, but it's something I will be able to
reuse for work, possibly completely replacing strongswan.
I have another Puppet module for IPsec which I was planning to
publish, but now I'm thinking I should just abort that and replace
everything with Wireguard, assuming we still need VPNs at work in the
future. (I have a number of reasons to believe we might not need any
in the near future anyways...)
Shantaram
I know I have been quite behind in review of books but then that s life. First up is actually not as much as a shocker but somewhat of a pleasant surprise. So, a bit of background before I share the news. If you have been living under a rock, then about 10-12 years ago a book called Shantaram was released. While the book is said to have been released in 2003/4 I got it in my hand around 2008/09 or somewhere around that. The book is like a good meal, a buffet. To share the synopsis, Lin a 20 something Australian guy gets involved with a girl, she encourages him to get into heroin, he becomes a heroin user. And drugs, especially hard drugs need constant replenishment, it is a chemical thing. So, to fund those cravings, he starts to steal, rising to rob a bank and while getting away shoots a cop who becomes dead. Now either he surrenders or is caught is unclear, but he is tortured in the jail. So one day, he escapes from prison, lands up at home of somebody who owes him a favor, gets some money, gets a fake passport and lands up in Mumbai/Bombay as it was then known. This is from where the actual story starts. And how a 6 foot something Australian guy relying on his street smartness and know how the transformation happens from Lin to Shantaram. Now what I have shared is perhaps just 5% of the synopsis, as shared the real story starts here.
Now the good news, last week 4 episodes of Shantaram were screened by Apple TV. Interestingly, I have seen quite a number people turning up to buy or get this book and also sharing it on Goodreads. Now there seems to have been some differences from the book to TV. Now I m relying on 10-12 year back memory but IIRC Khaderbhai, one of the main characters who sort of takes Lin/Shantaram under his wing is an Indian. In the series, he is a western or at least looks western/Middle Eastern to me. Also, they have tried to reproduce 1980s in Mumbai/Bombay but dunno how accurate they were My impression of that city from couple of visits at that point in time where they were still more tongas (horse-ridden carriages), an occasional two wheelers and not many three wheelers. Although, it was one of the more turbulent times as lot of agitation for worker rights were happening around that time and a lot of industrial action. Later that led to lot of closure of manufacturing in Bombay and it became more commercial. It would be interesting to know whether they shot it in actual India or just made a set somewhere in Australia, where it possibly might have been shot. The chawl of the book needs a bit of arid land and Australia has lots of it.
It is also interesting as this was a project that had who s who interested in it for a long time but somehow none of them was able to bring the project to fruition, the project seems to largely have an Australian cast as well as second generations of Indians growing in Australia. To take names, Amitabh Bacchan, Johnny Depp, Russel Crowe each of them wanted to make it into a feature film. In retrospect, it is good it was not into a movie, otherwise they would have to cut a lot of material and that perhaps wouldn t have been sufficient. Making it into a web series made sure they could have it in multiple seasons if people like it. There is a lot between now and 12 episodes to even guess till where it would leave you then. So, if you have not read the book and have some holidays coming up, can recommend it. The writing IIRC is easy and just flows. There is a bit of action but much more nuance in the book while in the web series they are naturally more about action. There is also quite a bit of philosophy between him and Kaderbhai and while the series touches upon it, it doesn t do justice but then again it is being commercially made.
Read the book, see the series and share your thoughts on what you think. It is possible that the series might go up or down but am sharing from where I see it, may do another at the end of the season, depending on where they leave it and my impressions.
Update A slight update from the last blog post. Seems Rishi Sunak seems would be made PM of UK. With Hunt as chancellor and Rishi Sunak, Austerity 2.0 seems complete. There have been numerous articles which share how austerity gives rises to fascism and vice-versa. History gives lot of lessons about the same. In Germany, when the economy was not good, it was all blamed on the Jews for number of years. This was the reason for rise of Hitler, and while it did go up by a bit, propaganda by him and his loyalists did the rest. And we know and have read about the Holocaust. Today quite a few Germans deny it or deny parts of it but that s how misinformation spreads. Also Hitler is looked now more as an aberration rather than something to do with the German soul. I am not gonna talk more as there is still lots to share and that actually perhaps requires its own blog post to do justice for the same.
The Pyramid by Henning Mankell
I had actually wanted to review this book but then the bomb called Shantaram appeared and I had to post it above. I had read two-three books before it, but most of them were about multiple beheadings and serial killers. Enough to put anybody into depression. I do not know if modern crime needs to show crime and desperation of and to such a level. Why I and most loved and continue to love Sherlock Holmes as most stories were not about gross violence but rather a homage to the art of deduction, which pretty much seems to be missing in modern crime thrillers rather than grotesque stuff.
In that, like a sort of fresh air I read/am reading the Pyramid by Henning Mankell. The book is about a character made by Monsieur Henning Mankell named Kurt Wallender. I am aware of the series called Wallender but haven t yet seen it. The book starts with Wallender as a beat cop around age 20 and on his first case. He is ambitious, wants to become a detective and has a narrow escape with death. I wouldn t go much into it as it basically gives you an idea of the character and how he thinks and what he does. He is more intuitive by nature and somewhat of a loner. Probably most detectives IRL are, dunno, have no clue. At least in the literary world it makes sense, in real world think there would be much irony for sure. This is speculation on my part, who knows.
Back to the book though. The book has 5 stories a sort of prequel one could say but also not entirely true. The first case starts when he is a beat cop in 1969 and he is just a beat cop. It is a kind of a prequel and a kind of an anthology as he covers from the first case to the 1990s where he is ending his career sort of.
Before I start sharing about the stories in the book, I found the foreword also quite interesting. It asks questions about the interplay of the role of welfare state and the Swedish democracy. Incidentally did watch couple of videos about a sort of mixed sort of political representation that happens in Sweden. It uses what is known as proportional representation. Ironically, Sweden made a turn to the far right this election season. The book was originally in Swedish and were translated to English by Ebba Segerberg and Laurie Thompson.
While all the stories are interesting, would share the last three or at least ask the questions of intrigue. Of course, to answer them you would need to read the book
So the last three stories I found the most intriguing.
The first one is titled Man on the Beach. Apparently, a gentleman goes to one of the beaches, a sort of lonely beach, hails a taxi and while returning suddenly dies. The Taxi driver showing good presence of mind takes it to hospital where the gentleman is declared dead on arrival. Unlike in India, he doesn t run away but goes to the cafeteria and waits there for the cops to arrive and take his statement. Now the man is in his early 40s and looks to be fit. Upon searching his pockets he is found to relatively well-off and later it turns out he owns a couple of shops. So then here are the questions ?
What was the man doing on a beach, in summer that beach is somewhat popular but other times not so much, so what was he doing there?
How did he die, was it a simple heart attack or something more? If he had been drugged or something then when and how?
These and more questions can be answered by reading the story Man on the Beach .
2. The death of a photographer Apparently, Kurt lives in a small town where almost all the residents have been served one way or the other by the town photographer. The man was polite and had worked for something like 40 odd years before he is killed/murdered. Apparently, he is murdered late at night. So here come the questions
a. The shop doesn t even stock any cameras and his cash box has cash. Further investigation reveals it is approximate to his average takeout for the day. So if it s not for cash, then what is the motive ?
b. The body was discovered by his cleaning staff who has worked for almost 20 years, 3 days a week. She has her own set of keys to come and clean the office? Did she give the keys to someone, if yes why?
c. Even after investigation, there is no scandal about the man, no other woman or any vices like gambling etc. that could rack up loans. Also, nobody seems to know him and yet take him for granted till he dies. The whole thing appears to be quite strange. Again, the answers lie in the book.
3. The Pyramid Kurt is sleeping one night when the telephone rings. The scene starts with a Piper Cherokee, a single piston aircraft flying low and dropping something somewhere or getting somebody from/on the coast of Sweden. It turns and after a while crashes. Kurt is called to investigate it. Turns out, the plane was supposed to be destroyed. On crash, both the pilot and the passenger are into pieces so only dental records can prove who they are. Same day or a day or two later, two seemingly ordinary somewhat elderly women, spinsters, by all accounts, live above the shop where they sell buttons and all kinds of sewing needs of the town. They seem middle-class. Later the charred bodies of the two sisters are found :(. So here come the questions
a.Did the plane drop something or pick something somebody up ? The Cherokee is a small plane so any plane field or something it could have landed up or if a place was somehow marked then could be dropped or picked up without actually landing.
b. The firefighter suspects arson started at multiple places with the use of petrol? The question is why would somebody wanna do that? The sisters don t seem to be wealthy and practically everybody has bought stuff from them. They weren t popular but weren t also unpopular.
c. Are the two crimes connected or unconnected? If connected, then how?
d. Most important question, why the title Pyramid is given to the story. Why does the author share the name Pyramid. Does he mean the same or the original thing? He could have named it triangle. Again, answers to all the above can be found in the book.
One thing I also became very aware of during reading the book that it is difficult to understand people s behavior and what they do. And this is without even any criminality involved in. Let s say for e.g. I die in some mysterious circumstances, the possibility of the police finding my actions in last days would be limited and this is when I have hearing loss. And this probably is more to do with how our minds are wired. And most people I know are much more privacy conscious/aware than I am.
Japan s Hikikomori
Japan has been a curious country. It was more or less a colonizer and somewhat of a feared power till it dragged the U.S. unnecessarily in World War 2. The result of the two atom bombs and the restitution meant that Japan had to build again from the ground up. It is also in a seismically unstable place as they have frequent earthquakes although the buildings are hardened/balanced to make sure that vibrations don t tear buildings apart. Had seen years ago on Natgeo a documentary that explains all that. Apart from that, Japan was helped by the Americans and there was good kinship between them till the 1980s till it signed the Plaza Accord which enhanced asset price bubbles that eventually burst. Something from which they are smarting even today. Japan has a constitutional monarchy. A somewhat history lesson or why it exists even today can be found here. Asset price bubbles of the 1980s, more than 50 percent of the population on zero hour contracts and the rest tend to suffer from overwork. There is a term called Karoshi that explains all. An Indian pig-pen would be two, two and a half times larger than a typical Japanese home. Most Japanese live in micro-apartments called konbachiku . All of the above stresses meant that lately many young Japanese people have become Hikikomori. Bloomberg featured about the same a couple of years back. I came to know about it as many Indians are given the idea of Japan being a successful country without knowing the ills and issues it faces. Even in that most women get the wrong end of the short stick i.e. even it they manage to find jobs, it would be most back-breaking menial work. The employment statistics of Japan s internal ministry tells its own story.
If you look at the data above, it seems that the between 2002 and 2019, the share of zero hour contracts has increased while regular work has decreased. This also means that those on the bottom of the ladder can no longer afford a home. There is and was a viral video called Lost in Manboo that went viral few years ago. It is a perfect set of storms. Add to that the Fukushima nuclear incident about which I had shared a few years ago. While the workers are blamed but all design decisions are taken by the management. And as was shown in numerous movies, documentaries etc. Interestingly, and somewhat ironically, the line workers knew the correct things to do and correct decisions to take unlike the management. The shut-ins story is almost a decade or two decades old. It is similar story in South Korea but not as depressive as the in Japan. It is somewhat depressive story but needed to be shared. The stories shared in the bloomberg article makes your heart ache
Backpacks
In and around 2015, I had bought a Targus backpack, very much similar to the Targus TSB194US-70 Motor 16-inch Backpack. That bag has given me a lot of comfort over the years but now has become frayed the zip sometimes work and sometimes doesn t. Unlike those days there are a bunch of companies now operating in India. There are eight different companies that I came to know about, Aircase, Harrisons Sirius, HP Oddyssey, Mokobara, Artic Hunter, Dell Pro Hybrid, Dell Roller Backpack and lastly the Decathlon Quechua Hiking backpack 32L NH Escape 500 . Now of all the above, two backpacks seem the best, the first one is Harrisons Sirius, with 45L capacity, I don t think I would need another bag at all. The runner-up is the Decathlon Quecha Hiking Backpack 32L. One of the better things in all the bags is that all have hidden pockets for easy taking in and out of passport while having being ant-theft. I do not have to stress how stressful it is to take out the passport and put it back in. Almost all the vendors have made sure that it is not a stress point anymore. The good thing about the Quecha is that they are giving 10 years warranty, the point to be asked is if that is does the warranty cover the zip. Zips are the first thing that goes out in bags.That actually has what happened to my current bag. Decathlon has a store in Wakad, Pune while I have reached out to the gentleman in charge of Harrisons India to see if they have a reseller in Pune. So hopefully, in next one week I should have a backpack that isn t spilling with things all over the place, whichever I m able to figure out.
Nona the Ninth is the third book of the Locked Tomb series and
entirely pointless to read if you have not read the series to date. It
completely spoils the previous books, assuming you would be able to figure
out who these people were and why you should care about them. This is
only for readers who are already invested.
This series was originally supposed to be a trilogy, and this book was
supposed to be Alecto the Ninth. Muir says in the acknowledgments,
and has said at more length elsewhere, that Nona changed all of her plans
and demanded her own book. Hence this book, postponing the end of the
series and lengthening it to four books.
After reading it, I understand why Muir decided to write a whole book
about Nona. She's an interesting character in ways that wouldn't have
come out if she was a small part of the concluding book. Unfortunately,
it's also obvious that this book wasn't part of the plan. It's not
entirely correct to say that Nona the Ninth is devoid of series
plot, but the plot advances very little, and mostly at the end.
Instead, we get Nona, who is physically a teenager who acts like someone
several years younger, most of the time. She lives with her family (who I
won't name to avoid spoilers for Harrow
the Ninth), helps at a local school (although her level of understanding
is about that of the students), and is a member of a kid's gang. She also
has dreams every night about a woman with a painted face, dreams her
family are very interested in.
This sounds weirdly normal for this series, but Nona and her family live
in a war-torn city full of fighting, refugees, and Blood of Eden
operatives. The previous books of the series took place in the rarefied
spaces of the Houses. Here we see a bit of the rest of the universe,
although it's not obvious at first what we're looking at and who these
people are. Absolutely no concession is made to the reader's fading
memory, so expect to need either a re-read, help from friends with better
memories, or quality time with a wiki. And, well, good luck with the
latter if you've not already read this book, since the
Locked
Tomb Wiki has now been updated with spoilers for Nona.
The other challenge, besides memory for the plot, is that this book is
told from a tight third-person focus on Nona, and Nona is not a very
reliable narrator. She doesn't lie, exactly, but she mostly doesn't
understand what's going on, often doesn't care, and tends not to focus on
what the reader is the most interested in. Nona is entirely uninterested
in developing the series plot. Her focus is on her child friends (who are
moderately interesting but not helpful if you're trying to figure out the
rest of the story) and the other rhythms of a strange life that's normal
to her.
For me at least, that meant the first half of this book involved a lot of
"what the heck is going on and why do I care about any of this?" I liked
Harrow the Ninth a lot, despite how odd and ambiguous it was, but I
was ready for revelations and plot coherence and was not thrilled by
additional complexity, odd allusions, and half-revealed details. I didn't
mind the layers of complexity added on by Harrow, but for me
Nona was a bit too much and I started getting frustrated rather
than intrigued.
We do, at last, get most of the history of this universe, including the
specific details of how John became God Emperor and how the Houses were
founded. That happens in odd interludes with a forced and somewhat
artificial writing style, but it's more straightforward and comprehensible
than I feared at first.
The pace of the story picks up considerably towards the end of the book,
finally providing the plot momentum that I was hoping for. Unfortunately,
it also gets more cryptic at the end of the book in ways that I didn't
enjoy. The epilogue, which is vital to understanding the climax of the
novel, took me three readings before I think I understood what happened.
If you preferred the clarity of Gideon
the Ninth, be warned that Nona is more like Harrow and
Muir seems to be making the plot more cryptic as she goes. I am hoping
this trend reverses in Alecto the Ninth.
This book made me grumpy. Nona is okay as a character, but the characters
in this series that I really like mostly do not appear or appear in
heavily damaged and depressing forms. Muir does bring back a couple of my
favorite characters, but then does something to them that's a major
spoiler but that I think was intended to be a wonderful moment for them
and instead left me completely cold and unhappy. There are still some
great moments of humor, but overall it felt more strained.
That said, I still had tons of fun discussing this book and its
implications with friends who were reading it at the same time. I think
that is the best way to read this series. Muir is being intentionally
confusing and is inserting a blizzard of references. Some of them are pop
culture jokes, but some of them are deep plot clues, and I'm not up to
deciphering them all by myself. Working through them with other people is
much more fun. (It also gives me an opportunity to feel smug about
guessing correctly what was happening at the end of Harrow the
Ninth, when I'm almost never the person who makes correct guesses about
that sort of thing.)
I think your opinion of this one will depend on how much you like Nona as
a character, how much patience you have for the postponement of plot
resolution, and how much tolerance you have for even more cryptic
references. I'm still invested in this series until the end, but this was
not my favorite installment. I suspect it (and the rest of the series)
would benefit immensely from re-reading, but life is short and my reading
backlog is long. What Muir is doing is interesting and has a lot of
depth, but she's asking quite a lot of the reader.
Content warning: Nona has an eating disorder, which occupied rather more
of my mental space while reading this book than I was comfortable with.
Followed by Alecto the Ninth, which does not have a publication
date scheduled as of this writing.
Rating: 7 out of 10
The purpose of this post is to demonstrate a first approach to the analysis of multiwavelength kinetic data, like those obtained using stopped-flow data. To practice, we will use data that were acquired during the stopped flow practicals of the MetBio summer school from the FrenchBIC. During the practicals, the student monitored the reaction of myoglobin (in its Fe(III) state) with azide, which yields a fast and strong change in the absorbance spectrum of the protein, which was monitored using a diode array. The data is publicly available on zenodo.
Aims of this tutorial
The purpose of this tutorial is to teach you to use the free softwareQSoas to run a simple, multiwavelength exponential fit on the data, and to look at the results. This is not a kinetics lecture, so that it will not go in depth about the use of the exponential fit and its meaning.
Getting started: loading the file
First, make sure you have a working version of QSoas, you can download them (for free) there. Then download the data files from zenodo. We will work only on the data file Azide-1.25mm_001.dat, but of course, the purpose of this tutorial is to enable you to work on all of them. The data files contain the time evolution of the absorbance for all wavelengths, in a matrix format, in which each row correpond to a time point and each column to a wavelength.
Start QSoas, and launch the command:
QSoas> load /comments='"'
Then, choose the Azide-1.25mm_001.dat data file. This should bring up a horizontal red line at the bottom of the data display, with X values between about 0 and 2.5. If you zoom on the red line with the mouse wheel, you'll realize it is data. The /comments='"' part is very important since it allows the extraction of the wavelength from the data. We will look at what it means another day. At this stage, you can look at the loaded data using the command:
QSoas> edit
You should have a window looking like this:
The rows each correspond to a data point displayed on the window below. The first column correspond to the X values, the second the Y values, and all the other ones to extra Y columns (they are not displayed by default). What is especially interesting is the first row, which contains a nan as the X value and what is obviously the wavelength for all the Y values. To tell that QSoas should take this line as the wavelength (which will be the perpendicular coordinate, the coordinate of the other direction of the matrix), first close the edit window and run:
QSoas> set-perp /from-row=0
Splitting and fitting
Now, we have a single dataset containing a lot of Y columns. We want to fit all of them simultaneously with a (mono) exponential fit. For that, we first need to split the big matrix into a series of X,Y datasets (because fitting only works on the first Y). This is possible by running:
QSoas> expand /style=red-to-blue /flags=kinetics
Your screen should now look like this:
You're looking at the kinetics at all wavelengths at the same time (this may take some time to display on your computer, it is after all a rather large number of data points). The /style=red-to-blue is not strictly necessary, but it gives the red to blue color gradient which makes things easier to look at (and cooler !). The /flags=kinetics is there to attach a label (a flag) to the newly created datasets so we can easily manipulate all of them at the same time. Then it's time to fit, with the following command:
QSoas> mfit-exponential-decay flagged:kinetics
This should bring up a new window. After resizing it, you should have something that looks like this:
The bottom of the fit window is taken by the parameters, each with two checkboxes on the right to set them fixed (i.e. not determined by the fitting mechanism) and/or global (i.e. with a single value for all the datasets, here all the wavelengths). The top shows the current dataset along with the corresponding fit (in green), and, below, the residuals. You can change the dataset by clicking on the horizontal arrows or using Ctrl+PgUp or Ctrl+PgDown (keep holding it to scan fast). See the Z = 728.15 showing that QSoas has recognized that the currently displayed dataset corresponds to the wavelength 728.15. The equation fitted to the data is: $$y(x) = A_\infty + A_1 \times \exp -(x - x_0)/\tau_1$$
In this case, while the \(A_1\) and \(A_\infty\) parameters clearly depend on the wavelength, the time constant of evolution should be independent of wavelength (the process happens at a certain rate regardless of the wavelength we're analyzing), so that the \(\tau_1\) parameter should be common for all the datasets/wavelengths. Just click on the global checkbox at the right of the tau_1 parameter, make sure it is checked, and hit the Fit button...
The fit should not take long (less than a minute), and then you end up with the results of the fits: all the parameters. The best way to look at the non global parameters like \(A_1\) and \(A_\infty\) is to use the Show Parameters item from the Parameters menu. Using it and clicking on A_inf too should give you a display like this one:
The A_inf parameter corresponds to the spectum at infinite time (of azide-bound heme), while the A_1 parameter corresponds to the difference spectrum between the initial (azide-free) and final (azide-bound) states.
Now, the fit is finished, you can save the parameters if you want to reload them in a later fit by using the Parameters/Save menu item or export them in a form more suitable for plotting using Parameters/Export (although QSoas can also display and the parameters saved using Save). This concludes this first approach to fitting the data. What you can do is
look at the depence of the tau_1 parameter as a function of the azide concentration;
try fitting more than one exponential, using for instance:
How to read the code above
All the lines starting by QSoas> in the code areas above are meant to be typed into the QSoas command line (at the bottom of the window), and started by pressing enter at the end. You must remove the QSoas> bit. The other lines (when applicable) show you the response of QSoas, in the terminal just above the command-line. You may want to play with the QSoas tutorial to learn more about how to interact with QSoas.
About QSoas
QSoas is a powerful open source data analysis program that focuses on flexibility and powerful fitting capacities. It is released under the GNU General Public License. It is described in Fourmond, Anal. Chem., 2016, 88 (10), pp 5050 5052. Current version is 3.1. You can freely (and at no cost) download its source code or precompiled versions for MacOS and Windows there. Alternatively, you can clone from the GitHub repository.
Contact: find my email address there, or contact me on LinkedIn.
I ve released a new tool to manage lockfiles for Arch Linux packages that can t use a lockfile from the official upstream release. It integrates closely with other Arch Linux tooling like updpkgsums that s already used to pin the content of build inputs in PKGBUILD.
To use this, the downstream lockfile becomes an additional source input in the source= array of our PKGBUILD (this is already the case for some packages).
You would then add a new function named updlockfiles that can generate new lockfiles and copies them into $outdir, and a prepare function to copy the lockfile in the right place:
To update the package to the latest (compatible) patch level simply run:
updlockfiles
This can also be used in case upstreams lockfile has vulnerable dependencies that you want to patch downstream. For more detailed instructions see the readme.
Thanks
This work is currently crowd-funded on github sponsors. I d like to thank @SantiagoTorres, @repi and @rgacogne for their support in particular.
I m on the train back from MuniHac, where I gave a talk about the rec-def library that I have excessively blogged about recently (here, here, here and here). I got quite flattering comments about that talk, so if you want to see if they were sincere, I suggest you watch the recording of Getting recursive definitions off their bottoms (but it s not necessary for the following).
After the talk, Franz Thoma approached me and told me a story of how we was once implementing the game Minesweeper in Haskell, and in particular the part of the logic where, after the user has uncovered a field, the game would automatically uncover all fields that are next to a neutral field, i.e. one with zero adjacent bombs. He was using a comonadic data structure, which makes a context-dependent parallel computation such as uncovering one field quite natural, and was hoping that using a suitable fix-point operator, he can elegantly obtain not just the next step, but directly the result of recursively uncovering all these fields. But, much to his disappointment, that did not work out: Due to the recursion inherent in that definition, a knot-tying fixed-point operator will lead to a cyclic definition.
Microsoft Minesweeper
He was wondering if the rec-def library could have helped him, and we sat down to find out, and this is the tale of this blog post. I will avoid the comonadic abstractions and program it more naively, though, to not lose too many readers along the way. Maybe read Chris Penner s blog post and Finch s functional pearl Getting a Quick Fix on Comonads if you are curious about that angle.
Minesweeper setup
Let s start with defining a suitable data type for the grid of the minesweeper board. I ll use the Array data type, it s Ix-based indexing is quite useful for grids:
Welcome to the September 2022 report from the Reproducible Builds project! In our reports we try to outline the most important things that we have been up to over the past month. As a quick recap, whilst anyone may inspect the source code of free software for malicious flaws, almost all software is distributed to end users as pre-compiled binaries. If you are interested in contributing to the project, please visit our Contribute page on our website.
David A. Wheeler reported to us that the US National Security Agency (NSA), Cybersecurity and Infrastructure Security Agency (CISA) and the Office of the Director of National Intelligence (ODNI) have released a document called Securing the Software Supply Chain: Recommended Practices Guide for Developers (PDF).
As David remarked in his post to our mailing list, it expressly recommends having reproducible builds as part of advanced recommended mitigations . The publication of this document has been accompanied by a press release.
Holger Levsen was made aware of a small Microsoft project called oss-reproducible. Part of, OSSGadget, a larger collection of tools for analyzing open source packages , the purpose of oss-reproducible is to:
analyze open source packages for reproducibility. We start with an existing package (for example, the NPM left-pad package, version 1.3.0), and we try to answer the question, Do the package contents authentically reflect the purported source code?
More details can be found in the README.md file within the code repository.
David A. Wheeler also pointed out that there are some potential upcoming changes to the OpenSSF Best Practices badge for open source software in relation to reproducibility. Whilst the badge programme has three certification levels ( passing , silver and gold ), the gold level includes the criterion that The project MUST have a reproducible build .
David reported that some projects have argued that this reproducibility criterion should be slightly relaxed as outlined in an issue on the best-practices-badge GitHub project. Essentially, though, the claim is that the reproducibility requirement doesn t make sense for projects that do not release built software, and that timestamp differences by themselves don t necessarily indicate malicious changes. Numerous pragmatic problems around excluding timestamps were raised in the discussion of the issue.
Sonatype, a pioneer of software supply chain management , issued a press release month to report that they had found:
[ ] a massive year-over-year increase in cyberattacks aimed at open source project ecosystems. According to early data from Sonatype s 8th annual State of the Software Supply Chain Report, which will be released in full this October, Sonatype has recorded an average 700% jump in repository attacks over the last three years.
More information is available in the press release.
A number of changes were made to the Reproducible Builds website and documentation this month, including Chris Lamb adding a redirect from /projects/ to /who/ in order to keep old or archived links working [], Jelle van der Waa added a Rust programming language example for SOURCE_DATE_EPOCH [][] and Mattia Rizzolo included Protocol Labs amongst our project-level sponsors [].
Debian
There was a large amount of reproducibility work taking place within Debian this month:
The nfft source package was removed from the archive, and now all packages in Debian bookworm now have a corresponding .buildinfo file. This can be confirmed and tracked on the associated page on thetests.reproducible-builds.org site.
Vagrant Cascadian announced on our mailing list an informal online sprint to help clear the huge backlog of reproducible builds patches submitted by performing NMU (Non-Maintainer Uploads). The first such sprint took place on September 22nd with the following results:
Holger Levsen:
Mailed #1010957 in man-db asking for an update and whether to remove the patch tag for now. This was subsequently removed and the maintainer started to address the issue.
Emailed #1017372 in plymouth and asked for the maintainer s opinion on the patch. This resulted in the maintainer improving Vagrant s original patch (and uploading it) as well as filing an issue upstream.
The plan is to repeat these sprints every two weeks, with the next taking place on Thursday October 6th at 16:00 UTC on the #debian-reproducible IRC channel.
Roland Clobus posted his 13th update of the status of reproducible Debian ISO images on our mailing list. During the last month, Roland ensured that the live images are now automatically fed to openQA for automated testing after they have been shown to be reproducible. Additionally Roland asked on the debian-devel mailing list about a way to determine the canonical timestamp of the Debian archive. []
Lastly, 44 reviews of Debian packages were added, 91 were updated and 17 were removed this month adding to our knowledge about identified issues. A number of issue types have been updated too, including the descriptions of cmake_rpath_contains_build_path [], nondeterministic_version_generated_by_python_param [] and timestamps_in_documentation_generated_by_org_mode []. Furthermore, two new issue types were created: build_path_used_to_determine_version_or_package_name [] and captures_build_path_via_cmake_variables [].
Other distributions
In openSUSE, Bernhard M. Wiedemann published his usual openSUSE monthly report.
diffoscopediffoscope is our in-depth and content-aware diff utility. Not only can it locate and diagnose reproducibility issues, it can provide human-readable diffs from many kinds of binary formats. This month, Chris Lamb prepared and uploaded versions 222 and 223 to Debian, as well as made the following changes:
The cbfstools utility is now provided in Debian via the coreboot-utils package so we can enable that functionality within Debian. []
Fixed the try.diffoscope.org service by addressing a compatibility issue between glibc/seccomp that was preventing the Docker-contained diffoscope instance from spawning any external processes whatsoever []. I also updated the requirements.txt file, as some of the specified packages were no longer available [][].
In addition Jelle van der Waa added support for file version 5.43 [] and Mattia Rizzolo updated the packaging:
Also include coreboot-utils in the Build-Depends and Test-Depends fields so that it is available for tests. []
Use pep517 and pip to load the requirements. []
Remove packages in Breaks/Replaces that have been obsoleted since the release of Debian bullseye. []
Reprotestreprotest is our end-user tool to build the same source code twice in widely and deliberate different environments, and checking whether the binaries produced by the builds have any differences. This month, reprotest version 0.7.22 was uploaded to Debian unstable by Holger Levsen, which included the following changes by Philip Hands:
Actually ensure that the setarch(8) utility can actually execute before including an architecture to test. []
Include all files matching *.*deb in the default artifact_pattern in order to archive all results of the build. []
Emit an error when building the Debian package if the Debian packaging version does not patch the Python version of reprotest. []
Remove an unneeded invocation of the head(1) utility. []
Upstream patches
The Reproducible Builds project detects, dissects and attempts to fix as many currently-unreproducible packages as possible. We endeavour to send all of our patches upstream where appropriate. This month, we wrote a large number of such patches, including:
Testing framework
The Reproducible Builds project runs a significant testing framework at tests.reproducible-builds.org in order to check packages and other artifacts for reproducibility. This month, however, the following changes were made:
Holger Levsen:
Add a job to build reprotest from Git [] and use the correct Git branch when building it [].
Mattia Rizzolo:
Enable syncing of results from building live Debian ISO images. []
Use scp -p in order to preserve modification times when syncing live ISO images. []
Apply the shellcheck shell script analysis tool. []
In a build node wrapper script, remove some debugging code which was messing up calling scp(1) correctly [] and consquently add support to use both scp -p and regular scp [].
Roland Clobus:
Track and handle the case where the Debian archive gets updated between two live image builds. []
Remove a call to sudo(1) as it is not (or no longer) required to delete old live-build results. []
Contact
As ever, if you are interested in contributing to the Reproducible Builds project, please visit our Contribute page on our website. However, you can get in touch with us via:
Canon Space is run, in a way, by the noble mercantile houses, who spread
their cities, colonies, and mines through the mysterious Web that allows
faster-than-light interstellar travel. The true rulers of Canon Space,
though, are the Guardships: enormous, undefeatable starships run by crews
who have become effectively immortal by repeated uploading and
reincarnation. Or, in the case of the Deified, without reincarnation,
observing, ordering, advising, and meddling from an entirely virtual
existence. The Guardships have enforced the status quo for four thousand
years.
House Tregesser thinks they have the means to break the stranglehold of
the Guardships. They have contact with Outsiders from beyond Canon Space
who can provide advanced technology. They have their own cloning
technology, which they use to create backup copies of their elites. And
they have Lupo Provik, a quietly brilliant schemer who has devoted his
life to destroying Guardships.
This book was so bad. A more sensible person than I would have given up
after the first hundred pages, but I was too stubborn. The stubbornness
did not pay off.
Sometimes I pick up an older SFF novel and I'm reminded of how much the
quality bar in the field has been raised over the past twenty years. It's
not entirely fair to treat The Dragon Never Sleeps as typical of
1980s science fiction: Cook is far better known for his Black
Company military fantasy series, this is one of his minor novels, and
it's only been intermittently in print. But I'm dubious this would have
been published at all today.
First, the writing is awful. It's choppy, cliched, awkward, and has no
rhythm or sense of beauty. Here's a nearly random paragraph near the
beginning of the book as a sample:
He hurled thunders and lightnings with renewed fury. The whole damned
universe was out to frustrate him. XII Fulminata! What the
hell? Was some malign force ranged against him?
That was his most secret fear. That somehow someone or something was
using him the way he used so many others.
(Yes, this is one of the main characters throwing a temper tantrum with a
special effects machine.)
In a book of 450 pages, there are 151 chapters, and most chapters switch
viewpoint characters. Most of them also end with a question or some
vaguely foreboding sentence to try to build tension, and while I'm willing
to admit that sometimes works when used sparingly, every three pages is
not sparingly.
This book is also weirdly empty of description for its size. We get a few
set pieces, a few battles, and a sentence or two of physical description
of most characters when they're first introduced, but it's astonishing how
little of a mental image I had of this universe after reading the whole
book. Cook probably describes a Guardship at some point in this book, but
if he does, it completely failed to stick in my memory. There are aliens
that everyone recognizes as aliens, so presumably they look different than
humans, but for most of them I have no idea how. Very belatedly we're
told one important species (which never gets a name) has a distinctive
smell. That's about it.
Instead, nearly the whole book is dialogue and scheming. It's clear that
Cook is intending to write a story of schemes and counter-schemes and
jousting between brilliant characters. This can work if the dialogue is
sufficiently sharp and snappy to carry the story. It is not.
"What mischief have you been up to, Kez Maefele?"
"Staying alive in a hostile universe."
"You've had more than your share of luck."
"Perhaps luck had nothing to do with it, WarAvocat. Till now."
"Luck has run out. The Ku Question has run its course. The symbol is
about to receive its final blow."
There are hundreds of pages of this sort of thing.
The setting is at least intriguing, if not stunningly original. There are
immortal warships oppressing human space, mysterious Outsiders, great
house politics, and an essentially immortal alien warrior who ends up
carrying most of the story. That's material for a good space opera if the
reader slowly learns the shape of the universe, its history, and its
landmarks and political factions. Or the author can decline to explain
any of that. I suppose that's also a choice.
Here are some things that you may have been curious about after reading
my summary, and which I'm still curious about after having finished the
book: What laws do the Guardships impose and what's the philosophy
behind those laws? How does the economic system work? Who built the
Guardships originally, and how? How do the humans outside of Canon Space
live? Who are the Ku? Why did they start fighting the humans? How
many other aliens are there? What do they think of this? How does the
Canon government work? How have the Guardships remained technologically
superior for four thousand years?
Even where the reader gets a partial explanation, such as what Web is and
how it was built, it's an unimportant aside that's largely devoid of a
sense of wonder. The one piece of world-building that this book is
interested in is the individual Guardships and the different ways in which
they've handled millennia of self-contained patrol, and even there we only
get to see a few of them.
There is a plot with appropriately epic scope, but even that is undermined
by the odd pacing. Five, ten, or fifty years sometimes goes by in a
sentence. A war starts, with apparently enormous implications for Canon
Space, and then we learn that it continues for years without warranting
narrative comment. This is done without transitions and without signposts
for the reader; it's just another sentence in the narration, mixed in with
the rhetorical questions and clumsy foreshadowing.
I would like to tell you that at least the book has a satisfying ending
that resolves the plot conflict that it finally reveals to the reader, but
I had a hard time understanding why the ending even mattered. The plot
was so difficult to follow that I'm sure I missed something, but it's not
difficult to follow in the fun way that would make me want to re-read it.
It's difficult to follow because Cook doesn't seem able to explain the
plot in his head to the reader in any coherent form. I think the status
quo was slightly disrupted? Maybe? Also, I no longer care.
Oh, and there's a gene-engineered sex slave in this book, who various
male characters are very protective and possessive of, who never develops
much of a personality, and who has no noticeable impact on the plot despite
being a major character. Yay.
This was one of the worst books I've read in a long time. In retrospect,
it was an awful place to start with Glen Cook. Hopefully his better-known
works are also better-written, but I can't say I feel that inspired to
find out.
Rating: 2 out of 10
I have released version 1.0.0 of Hippotat, my IP-over-HTTP system. To quote the README:
You re in a cafe or a hotel, trying to use the provided wifi. But it s not working. You discover that port 80 and port 443 are open, but the wifi forbids all other traffic.
Never mind, start up your hippotat client. Now you have connectivity. Your VPN and SSH and so on run over Hippotat. The result is not very efficient, but it does work.
Story
In early 2017 I was in a mountaintop cafeteria, hoping to do some work on my laptop. (For Reasons I couldn t go skiing that day.) I found that local wifi was badly broken: It had a severe port block. I had to use my port 443 SSH server to get anywhere. My usual arrangements punt everything over my VPN, which uses UDP of course, and I had to bodge several things. Using a web browser directly only the wifi worked normally, of course - otherwise the other guests would have complained. This was not the first experience like this I d had, but this time I had nothing much else to do but fix it.
In a few furious hacking sessions, I wrote Hippotat, a tool for making my traffic look enough like ordinary web browsing that it gets through most stupid firewalls. That Python version of Hippotat served me well for many years, despite being rather shonky, extremely inefficient in CPU (and therefore battery) terms and not very productised.
But recently things have started to go wrong. I was using Twisted Python and there was what I think must be some kind of buffer handling bug, which started happening when I upgraded the OS (getting newer versions of Python and the Twisted libraries). The Hippotat code, and the Twisted APIs, were quite convoluted, and I didn t fancy debugging it.
So last year I rewrote it in Rust. The new Rust client did very well against my existing servers. To my shame, I didn t get around to releasing it.
However, more recently I upgraded the server hosts my Hippotat daemons run on to recent Debian releases. They started to be affected by the bug too, rendering my Rust client unuseable. I decided I had to deploy the Rust server code.
This involved some packaging work. Having done that, it s time to release it: Hippotat 1.0.0 is out.
The package build instructions are rather strange
My usual approach to releasing something like this would be to provide a git repository containing a proper Debian source package. I might also build binaries, using sbuild, and I would consider actually uploading to Debian.
However, despite me taking a fairly conservative approach to adding dependencies to Hippotat, still a couple of the (not very unusual) Rust packages that Hippotat depends on are not in Debian. Last year I considered tackling this head-on, but I got derailed by difficulties with Rust packaging in Debian.
Furthermore, the version of the Rust compiler itself in Debian stable is incapable of dealing with recent versions of very many upstream Rust packages, because many packages most recent versions now require the 2021 Edition of Rust. Sadly, Rust s package manager, cargo, has no mechanism for trying to choose dependency versions that are actually compatible with the available compiler; efforts to solve this problem have still not borne the needed fruit.
The result is that, in practice, currently Hippotat has to be built with (a) a reasonably recent Rust toolchain such as found in Debian unstable or obtained from Rust upstream; (b) dependencies obtained from the upstream Rust repository.
At least things aren t completely terrible: Rustup itself, despite its alarming install rune, has a pretty good story around integrity, release key management and so on. And with the right build rune, cargo will check not just the versions, but the precise content hashes, of the dependencies to be obtained from crates.io, against the information I provide in the Cargo.lock file. So at least when you build it you can be sure that the dependencies you re getting are the same ones I used myself when I built and tested Hippotat. And there s only 147 of them (counting indirect dependencies too), so what could possibly go wrong?
Sadly the resulting package build system cannot work with Debian s best tool for doing clean and controlled builds, sbuild. Under the circumstances, I don t feel I want to publish any binaries.
Rama II
This would be more of a short post about the current book I am reading. Now people who have seen Arrival would probably be more at home. People who have also seen Avatar would also be familiar to the theme or concept I am sharing about. Now before I go into detail, it seems that Arthur C. Clarke wanted to use a powerful god or mythological character for the name and that is somehow the RAMA series started.
Now the first book in the series explores an extraterrestrial spaceship that earth people see/connect with. The spaceship is going somewhere and is doing an Earth flyby so humans don t have much time to explore the spaceship and it is difficult to figure out how the spaceship worked. The spaceship is around 40 km. long. They don t meet any living Ramans but mostly automated systems and something called biots.
As I m still reading it, I can t really say what happens next. Although in Rama or Rama I, the powers that be want to destroy it while in the end last they don t. Whether they could have destroyed it or not would be whole another argument. What people need to realize is that the book is a giant What IF scenario.
Aliens
If there were any intelligent life in the Universe, I don t think they will take the pain of visiting Earth. And the reasons are far more mundane than anything else. Look at how we treat each other. One of the largest democracies on Earth, The U.S. has been so divided. While the progressives have made some good policies, the Republicans are into political stunts, consider the political stunt of sending Refugees to Martha s Vineyard. The ex-president also made a statement that he can declassify anything just by thinking about it. Now understand this, a refugee is a legal migrant whose papers would be looked into by the American Govt. and till the time he/she/their application is approved or declined they can work, have a house, or do whatever to support themselves. There is a huge difference between having refugee status and being an undocumented migrant. And it isn t as if the Republicans don t know this, they did it because they thought they will be able to get away with it.
Both the above episodes don t throw us in a good light. If we treat others like the above, how can we expect to be treated? And refugees always have a hard time, not just in the U.S, , the UK you name it. The UK just some months ago announced a controversial deal where they will send Refugees to Rwanda while their refugee application is accepted or denied, most of them would be denied.
The Indian Government is more of the same. A friend, a casual acquaintance Nishant Shah shared the same issues as I had shared a few weeks back even though he s an NRI. So, it seems we are incapable of helping ourselves as well as helping others. On top of it, we have the temerity of using the word alien for them.
Now, just for a moment, imagine you are an intelligent life form. An intelligent life-form that could coax energy from the stars, why would you come to Earth, where the people at large have already destroyed more than half of the atmosphere and still arguing about it with the other half. On top of it, we see a list of authoritarian figures like Putin, Xi Jinping whose whole idea is to hold on to power for as long as they can, damn the consequences. Mr. Modi is no different, he is the dumbest of the lot and that s saying something. Most of the projects made by him are in disarray, Pune Metro, my city giving an example. And this is when Pune was the first applicant to apply for a Metro. Just like the UK, India too has tanked the economy under his guidance. Every time they come closer to target dates, the targets are put far into the future, for e.g. now they have said 2040 for a good economy. And just like in other countries, he has some following even though he has a record of failure in every sector of the economy, education, and defense, the list is endless. There isn t a single accomplishment by him other than screwing with other religions. Most of my countrymen also don t really care or have a bother to see how the economy grows and how exports play a crucial part otherwise they would be more alert. Also, just like the UK, India too gave tax cuts to the wealthy, most people don t understand how economies function and the PM doesn t care. The media too is subservient and because nobody asks the questions, nobody seems to be accountable :(.
Religion
There is another aspect that also has been to the fore, just like in medieval times, I see a great fervor for religion happening here, especially since the pandemic and people are much more insecure than ever before. Before, I used to think that insecurity and religious appeal only happen in the uneducated, and I was wrong. I have friends who are highly educated and yet still are blinded by religion. In many such cases or situations, I find their faith to be a sham. If you have faith, then there shouldn t be any room for doubt or insecurity. And if you are not in doubt or insecure, you won t need to talk about your religion. The difference between the two is that a person is satiated himself/herself/themselves with thirst and hunger. That person would be in a relaxed mode while the other person would continue to create drama as there is no peace in their heart.
Another fact is none of the major religions, whether it is Christianity, Islam, Buddhism or even Hinduism has allowed for the existence of extraterrestrials. We have already labeled them as aliens even before meeting them & just our imagination. And more often than not, we end up killing them. There are and have been scores of movies that have explored the idea. Independence day, Aliens, Arrival, the list goes on and on. And because our religions have never thought about the idea of ET s and how they will affect us, if ET s do come, all the religions and religious practices would panic and die. That is the possibility why even the 1947 Roswell Incident has been covered up .
If the above was not enough, the bombing of Hiroshima and Nagasaki by the Americans would always be a black mark against humanity. From the alien perspective, if you look at the technology that they have vis-a-vis what we have, they will probably think of us as spoilt babies and they wouldn t be wrong. Spoilt babies with nuclear weapons are not exactly a healthy mix
Earth
To add to our fragile ego, we didn t even leave earth even though we have made sure we exploit it as much as we can. We even made the anthropocentric or homocentric view that makes man the apex animal and to top it we have this weird idea that extraterrestrials come here or will invade for water. A species that knows how to get energy out of stars but cannot make a little of H2O. The idea belies logic and again has been done to death. Why we as humans are so insecure even though we have been given so much I fail to understand. I have shared on numerous times the Kardeshev Scale on this blog itself.
The above are some of the reasons why Arthur C. Clarke s works are so controversial and this is when I haven t even read the whole book. It forces us to ask questions that we normally would never think about. And I have to repeat that when these books were published for the first time, they were new ideas. All the movies, from Stanley Kubrick s 2001: Space Odyssey, Aliens, Arrival, and Avatar, somewhere or the other reference some aspect of this work. It is highly possible that I may read and re-read the book couple of times before beginning the next one. There is also quite a bit of human drama, but then that is to be expected. I have to admit I did have some nice dreams after reading just the first few pages, imagining being given the opportunity to experience an Extraterrestrial spaceship that is beyond our wildest dreams. While the Governments may try to cover up or something, the ones who get to experience that spacecraft would be unimaginable. And if they were able to share the pictures or a Livestream, it would be nothing short of amazing.
For those who want to, there is a lot going on with the New James Webb Telescope. I am sure it would give rise to more questions than answers.