Sami Haahtinen: Debian Xen dom0 Upgrade woes
I finally decided that it's time for me to upgrade my Xen
installation. It used to run etch with backported Xen, because the
etch version was increasingly difficult to work with.
I also acknowledge that some of the issues I've been having are
simply caused by yours truly, but even still the Debian Xen
installation is way too fragile to my taste. I've already
considered installing XenServer Express locally and running the
hosts on it. The big drawback has been that XenCenter (the tool
that is used to manage XenServer) is windows only and it doesn't
work with wine.
So you can imagine my desperation...
Anyway, the latest upgrade from etch to lenny was painful as
usual. The first part went smoothly, bit of sed magic on
sources.list and a few upgrade commands (carefully picking the Xen
packages out of the upgrade set). So in the end I had a working
lenny installation with backported Xen.
Next I made sure that there was nothing major going on in my
network (one of the virtual machines acts as my local firewall) and
took a deep breath before upgrading the rest of the packages. I
knew to be careful about xendomains -script which has reliably
restored my virtual machines after reboot to a broken host so I had
always ended up restarting my virtual machines after reboot.
I carefully cleared
XENDOMAINS_AUTO and set
XENDOMAINS_RESTORE to false in
/etc/default/xendomains so that the virtual machines
would be saved but not restored or restarted on reboot.
After the normal pre-boot checks I went for it.
Oddly enough everything worked normally and the system came up
after a bit of waiting. I checked the bridges and everything
appeared normal, so it was time to try and restore a single domain
to see that everything actually did work as planned.
Hydrogen:~# xm restore /var/lib/xen/save/Aluminium
Error: Device 0 (vif) could not be connected. Hotplug scripts not working.
Oof, Googling for the issue revealed that there were others that
had suffered from the same problem on various different platforms
the problems were caused by different things. One would assume that
the problem is in the vif-bridge script that is
mentioned in the xend-config.sxp file as the script
that brings up the vif, but after many hours of tial and error and
pointless googling (over gprs connection), I couldn't find any
solution to the problem. It was time to call it a day (it was
almost 3 am already...)
During the night I had a new idea about the possible cause. What
if the problem isn't in xend, but somewhere else. I fired up
udevadm monitor to see what udev saw and it wasn't
much. I'm not an expert with udev, but from previous encounters I
had a vague feeling that there was supposed to be more events
flying around.
I wasn't able to pinpoint what was wrong so I decided to purge
xen-utils, of which I had 2 versions installed: 3.2-1 and 3.0.2. I
also removed everything related to xenstore. After reinstalling the
current versions and restoring my configuration files the first
host came up just fine.
I still had problems resuming the virtual machines and I ended
up rebooting them again, which was nothing new, but at least they
were running again.
In the end I don't know what was the actual cause for udev not
handling the devices properly, but I'm happy to have them all
running again. And I learned a valuable lesson of all this: udev is
an important part of Xen, make sure it works properly.
I've been converting all my Mercurial repositories to Git. One of the
motivations for this was hg's poor handling of branches and tags:
branches are just another metadata field on a commit, and tags are
entries in a text file called .hgtags that is tracked in the
repository(!). Git has its flaws as well, but I prefer its view that
branches and tags are just refs, which are pointers that exist at the
repository level and are pushed or pulled around in much the same way
as the objects they refer to.
The excellent hg-to-git.py script included in contrib created Git tags
corresponding to my old Mercurial tags, but of course didn't modify
the actual commits. So I still had a .hgtags file in my Git
repository, and a boilerplate commit of the form "Add tag foo for
changeset bar" each time I added to it (recall that .hgtags is just
another file. Thankfully, I never had to deal with two branches where
I added different tags...). I wanted to remove these. Git has a
'filter-branch' command that can totally rewrite or expunge commits;
this is of course a horrible idea for already published code, but
there's no harm in using while initially preparing a repository.
While I appreciate git's object model and speed, I must agree with its
detractors that the user interface is terrible. It took some tinkering
to get git-filter-branch to do what I wanted, so I'm writing this to
save my notes for next time (and in case someone else is searching for
how to do this). Here's the command I arrived at:
I have a bit of money invested in stocks and funds, passively
managed by my bank. I ve made a bit with risky speculations, and
I ve lost about the same amount with the markets dive of the last
18 months. Net benefit: zero cash, a bit of experience. Time to
move on
I am not an interesting customer, so I was all the more
surprised to get a call from an employee of the bank the other day.
He identified himself as Crisis Manager or something similarly
ridiculous, mentioned at least 20 times that he was a regular
employee and wouldn t receive provisions, and he also kept assuring
me that he s doing himself what he s suggesting to me: leveraging
the cost-average effect.
I turned him down, probably a bit too briskily. Let s
investigate why.
Banks are good at devising ingenious tricks to fool people; it s
their business after all. One of those tricks is the so-called
cost-average effect. It s awesome, they say, because it is the
first product that adapts to the ever-changing market: when prices
are up, it buys smaller numbers of shares, and in low times, it
stocks up on them by buying larger quantities, automatically.
Fancy, no? It isn t, because the entire idea is to get your
signature on a contract that obliges you to pay a fixed sum each
month, with which shares are then bought. Obviously, when prices
are up, numbers you get for your money are low, and vice versa.
Observe how well the banks manage to turn that basic rule into a
selling point. They are good at that. In fact, that s probably all
they re good at. Obviously , inflation is also factored in, so the
monthly fixed sum actually increases by a projected 5% each year,
which they call inflation protection .
While this strategy investing fixed-sum installments rather
than lump sums certainly has its benefits, and can level out the
risk of throwing a large sum of money at the market, it is commonly
considered a suboptimal strategy, especially when employed over
longer spans of time. For more information, I refer you to the
frequently cited article
At some point in the past, when I was still thinking that
I've been a bit surprised by the strong positive response to my
Here's an idea: what if there was a website where you could upload a topical blog post, which would then be reviewed and edited by experts in the area and then posted on that website? Such a process would ensure that articles are at least well-named, appropriately tagged, well-structured and not methodologically deficient, and could result in a database of articles of decent quality. Would you upload your blog posts that are juicier in terms of content to such a site? Would you volunteer to do reviews and editing?

The changes in
This blog post is long overdue, but like Lenny, it was finally released. Yes, I defended my PhD thesis on December 4th. So I m officially a Doctor from the Universit Joseph Fourier Grenoble.
The general context of my thesis is distributed systems, like peer-to-peer systems and applications running on HPC (High Performance Computing) cluster and grids. More specifically, I worked on methods and tools to study such systems, trying to answer the following questions: take a given application that will run on a distributed system. What can we learn or verify about its performance or its characteristics? Will it scale well, even with 1000 or 10000 nodes? Does this app perform better than this another app? Under which conditions? (that s probably the hardest question, because you need to be able to run both apps under the same conditions if you want a fair comparison!)
My work involved studying basic building blocks that can be used to study distributed systems, like network emulators (Dummynet, Linux s traffic control subsystem, NISTNet). I also designed P2PLab, which allows to emulate network-centric peer-to-peer systems using a cluster, targetting experiments with thousands of emulated instances of the application. For example, it allowed to run experiments with ~15000 emulated BitTorrent nodes using 160 real machines.
So what s next? Well, my first-priority plan is to make use of my PhD to apply to an academic position (maitre de conf rences, i.e assistant professor). If you know an open position where I would be a good fit, don t hesitate to tell me! Of course, this plan might very well fail (I will know around end of June). In that case, I ll probably be looking for a job, likely Free-Software-related. Again, if you hear of something for me, ping me!


I've prepared a private autobuilder. As I've setup the incoming using scp, I don't need to gpg sign each package. But dupload complained saying that the package is not gpg signed. Lurking in dupload code, I found It's possible overriding the default preupload for my default host:
Ladies and gentlemen, it seems that Google has finally done it: they shipped Multics for 386 (and
compatible).
Have you ever thought what you send in email footer?
Yeah, yeah, I may be living under a rock these days. Anyway, I just
noticed that the GFDL v1.3 is