Sunday

The conference is held in the University Arms hotel, which is down the road (towards the trainstation) from where we are staying.
It only takes 10 minutes there on feet, even nif we take a detour (tested! ;)

After checkin, we located Room A for the first lecture. I tried to write a report of the talk while I was listening to it, which isn't easy at all... So what if I'm not reporter material ;) At least I can still bore all you people reading this !

The Kernel Report - Jonathan Corbet (LWN.net)
[G2:11956][G2:11960]
I'm writing this post offlinne because there is no wireless, or i don't knonw how to active it yet.
The talk I'm in right now is called the kernel report. It gives an overview of all new features (and future features) of the 2.6.23 kernel, which is about to be released.
(There has been some work on userspace drivers, which would be cool for things like nvidia drivers and such)

On a sidenote, I was reading the "linux network internals" book from O'Reilly yesterday and this morning and was wondering if I had a copy of the kernel lying around.
I didn't. Luckily, the conference CD that I received in the welcome pack has a git tree on it with the linux 2.6.23 kernel code.

Anyway, back to the talk.

In the filesystems area, I should have a look at the brtfs filesystem for various reasons (e.g. it can do snapshotting, checksumming and does fast fsck). I think it would be interesting to use on the host system of some virtual machines.
Some work on virtualisation is also being doing. Most notably, work by Rusty Russell called Lguest, which is a fully functional but very simple virtualisation system.

Since I'm typing this while the talk is going, I'm missing some of the details. But the speaker advises to install powertop to check out which processes are making my laptop come out of the idle-state and sucking more power through my laptop. This sounds interesting because right now, my laptop is not hooked up to the powergrid.
(Even worse, the power adapter I bought doesn't work here, so I cant recharge my battery using it. Luckily, my colleague has one that does work :)

Dag wieers asked what the deal is with Xen being trademarked. I didn't hear about that, but I don't like Xen anyway, so its good news :P

[Update: I tried installing powertop when I got network, but it isn't in my apt repository yet]

OS Circular on QEMU/KQEMU/KVM/Xen-HVM - Kuniyasu Suzaki (National Institute of Advanced Industrial Science and Techno)

[G2:11963]
(Note the chinese version of Ubuntu ;)

It seems OS Circular is actually a software system, and doesn't mean "OS overview".

Sitting in the back, I'm not making out much of what the speaker is saying. the airco is behind me making some noise, the speaker is asian (complete with accent) and speaks softly. So this talk is probably not going to be very interesting.

Luckily, we received the password for the wireless accesspoint in the previous talk, so I can go online and do some things like install ctags, which I don't yet have apparently.

After 10 minutes or so, we left this talk to make room for other people who really wanted to be there. We went to the O'Reilly bookstand and I bought 2 books: "Understanding the Linux Kernel" by Daniel P. Bovet and Marco Cesati (ISBN-13: 978-0-596-00565-8) and "Linux Kernel in a Nutshell" by Greg Kroah-Hartman (ISBN-13: 978-0-596-10079-7). The lady at the stand also included a poster with a genealogy overview of programming languages, and a "Developer Notebook", which is just a notepad, but with printed coffeestains on it ;)

TCP congestion testing: preventing global internet warming - Stephen Hemminnger (Linux Foundation)

I was somewhat looking forward to this talk because it was related to the networking field, but the speaker didn't show up. So we went to the lounge and browsed our books a bit, untill I noticed Dag Wieers who was there also.

Dag is set to give a talk on dstat, a tool he wrote, on monday.

Ganeti: an opensource multi-node HA cluster based on Xen
Instead of going to this talk, I stayed with Dag and learned more about dstat. The tool can be used to track resources in realtime. Instead of using watch to keep an eye on how much memory of diskspace is in use, dstat presents all this information on a single line and updates it at a given interval.
The tool made me think back to multiping, a small tool I wrote some time ago to monitor several machines at once. It would learn the state of all listed machines (whether they were up or down) and then poll them every second and report changes in the state. I later extended multiping by not just using ping anymore, but any program. That way, I can easily test if the LDAP-servers are still responding to queries and the SSL-certificates are still valid and such, all on a huge number of hosts.

Dag created the dstat tool in Python to help him with his new company (Dagit). It's very modular and should be able to monitor pretty much anything, provided you write a plugin for it. I made a couple suggestions, like using a namescheme "module.variable_name" to specify what data to show from each module (you can customize what dstat prints to screen) and allowing the user to set a range indicating what "normal behaviour" means for a certain variable. Using that range, dstat could report any values that fall outside of normal behaviour and the values cann be shown in red (so you can see something is wrong from across a room) and maybe give an audio signal (in case you're not looking to your screen)

Lunch
[G2:12050]

Lunch was a buffet, but it appeared badly organised. I had an interesting talk about the RedHat licenses and CentOS, with Dag. RedHat doesn't want to acknowledge CentOS and offer support for it (CentOS is exactly the same as the RedHat commercial linux distributions, but it's compiled from source by the community, and there is no commercial support). I understand why RedHat doesn't acknowledge it, since CentOS pulls away potential clients from RedHat commercial distributions.

The whole conversation left me with a bad aftertaste. Not because of the conversation itself, but because I realised more than ever why I like free software and especially the "free" part in it. No licenses, no copyrights, non lawyers and marketing departments. Give me Free(dom| software) or give me death ;) Still loving Debian and Ubuntu.

OpenVZ and Linux Virtualization - Kir Kolyshkin (OpenVZ)
[G2:12069]

OpenVZ is a container virtualization system like Vservers. It wasn't what I'm looking for. I'm looking for a completely shielded virtualised guest system, which can run any operating system and is FREE. At this point, I have (afaik) the choice between VMWare, Virtualbox, QEMU and KVM. VMWare is closed source. Virtualbox is opensource for the most part andn has Ubuntu packages (which work greatly btw, although the networking setup sucks). QEMU works too, but its kindof slow. KVM is not supported by my processor.

Automated system and service monitoring with OpenQRM and Nagios - Matthias Rechenburg (Freelancer/Project Manager of the OpenQRM project)
[G2:12073]

OpenQRM can manage a pool of resources in a datacenter. That description is pretty vague, because I wasn't really paying attention ;)
Maybe I should have though, because it looked to me that OpenQRM can be very useful in combination with Nagios. When Nagios detects a failure somewhere, it can report that to OpenQRM, which allocates a server from a pool, loads a new software image on it and uses that server to resolve the failure (for example in a cluster).

At this point, I wanted to leave because none of the remaining topics were of real interest to me. But my colleague wanted to see the DRBD8 talk, so we stayed a bit.

DRBD8: a shared disk semantics on a shared nothing clusterBofB - Lars Ellenberg (LINBIT)
[G2:12076]

I saw this same talk (or more or less the same) at LinuxKongress some time ago, but it was about version 0.7. Version 8 is the next release and promised active/active setups. At the time, I was considering using DRBD as a shared disk between 2 machines in a small cluster, to do failover of a single logical system/server. The data would be shared between nboth machines. If one went down, the other could takeover without interruption.

But the setup I had in mind (with OpenSSI), was too complicated and required too much work, so it was dropped.

It's good to know that DRBD is still moving forward.



The day was supposed to end with a fancy social event, but we didn't register for that. Neither of us felt like wearing fancy clothes for it ;) (Or actually: neither of us wanted to pack fancy clothes in our luggage)

Instead of the social event, we went to the Gourmet Burger Kitchen and ate there.
[G2:12039]