Those interested by the previous part of the story can read the previous article on LCA14.
I really had to sleep a bit that night (5 hours is a minimum these days) so I missed the lightning talks of the morning, but it doesn’t seem I missed a major announce there. But I didn’t want to miss the start of the main conferences.
As a followup of the previous day, and in order to know better what we, HP, do for the OpenStack project, I chose to attend Elisabeth Krumbach Joseph who was describing the management of the project infrastructure. She made a good description of the tools used, including some navigation on the various Web interfaces they are using. She insisted on the fact that all the infrssctructure is managed the same way as the rest of the OpenSTack project, which means in the open, and using gated trunk as well ,which is much less usual for an infrstucture than for a project. It sounded to me as a risk of adding too much bureaucracy, and was maybe more suited for an established infra not moving too much, but they seem to deal with it without problem. Definitely an interesting approach that to have peer review of infrastructure configuration changes before applying them. Maybe my Solution Center could get some inspiration out of it. Will see when I’m back. The ~6 people are managing around 35 servers for the OpenStack infra (all virtual, so they do not deal with HW issues).
As she had time left, I think nxt time she could demo in more details some of the tools she mentioned, and give a bit more details around sizing (which I gathered from questions I asked).
As I’m always interested by build or execution environment, I then attended the presentation from Jay Coles on Linception: playing with conatiners under Linux. It took me nearly the full presentation to understand what it really was about, as I first thought it would be an explanation of LXC or something similar, but it was in fact again a new implementation of containers for Linux, using various capabilities provided by the kernel. (Thanks to the audience member who asked for a demo at the, which made it completely clear to me then !). And re-reading the description of the talk, I think the content wasn’t completely what was described. Anyway, was interesting, but do we really need again a new solution in that area, after LXC, Libvirt-lxc and systemd-spawn ? Look yourself at http://www.doger.io/ to make your own idea. I’m not completely convinced.
After lunch, Mark Nottingham explained to us what HTTP 2.0 was all about. Plenty of good reasons to improve HTTP for sure, with lots of graphs demonstrating the issues. But guys, please stop making stuff so administrator unfriendly ! Compressing header may be a win for resources, but for sure a loss for us. We won’t be able to make a simple
telnet host 80 and type
GET / HTTP/1.0 e.g to check rapidly stuff. I don't think that helps at the end. Of course modifications are required to solve real issues. But be smart and invent something more friendly for us.
Then was a conference from someone I really wanted to see, if only for his vcs-pkg.org initiative. And more over because Martin Krafft was talking about configuration management for system administration, which is something I'm also working on for my systems, even if I don't have the final solution at the moment (ansible looking really close to what I like the most). And his presentation was in the line of what I expected: he prefers pets to cattle, and I must confess so do I ;-) He has strong opinion (he is a Debian advocate ;-)) on system management, and most of them sound pretty acurate to me. Martin developed a tool called reclass to help him encapsulate Salt and Ansible and do what he wants them to do. Definitely worth exploring. Even if it seems that latest versions of Salt have similar features, making the tool less useful. Anyway, I prefer ansible (for the SSH communication aspect), so added to my TODO list, which never gets empty !
After that talk, I could not miss the only one made by a french hacker ! Especially as I know Thomas Petazzoni for a long time now, have loved his Kernel presentations he was making in the past, and was interested to see a new tool such as BuildRoot, which could be relevant in the MondoRescue context. And it seems I wasn't the only one as the room was pretty packed for his presentation, which is good for his first LCA.
Is it chauvinism ? I found his presentation one of the best of the week. He gave a very detailed view on the tool features, giving concrete example of usage corresponding to the various configurations he was describing. But all that remaining cristal clear. The presentation was easily understood by the audience. The tool itself is pretty impressive, recompiling in order all what is needed to make a standalone Linux from scratch system ready to be embedded on any system. Of course, Thomsa knows the subject extremely well as he does that for a living for free electron, which was appearing on Jon Corbets's stats as a major ARM kernel contributors company.
He even mentioned the help that is provided with regards to licenses of the software contained in the build, which is a topic too few developers take seriously, and it was great to see him mentioning that.
My only regret: I'd have loved to have a short demo of on of his existing system, but honestly he had no additional time to do it, so could be worth adding for a one hour time slot. Anyway, well done Thomas, and looking forward to attend more of your presentation, as I always learn stuff when I do, which makes my day :-)
And finally, I finished the day with another of my favorite speakers in such event, Ric Wheeler from Red Hat who was talking about File systems and storage systems of course ! But this time he was exploring the concerns that bring to the Linux Kernel community the new storage systems arriving: Flash components used on memory slots, post-flash components, with low lantency and high storage capabilities (such HP's own memristor e.g.). But even if it creates some difficulties to the kernel hackers to keep with the pace of storage technology, these are really good news, as first storage technology will now also make a giant leap, as was done for network, CPU in the past. We will have TB of data in a very small form factor soon, providing bandwidth and latency optimal accesses. Then the Linux community is part of the developments made around these new devices, instead of running behind as it was the case in its enfancy. And finally it will change a lot the way we architect solutions in the future, especially thinking about optic introduction to interconnect all these new components. So the future 10 years are a as bright as were the past 20 ones at least.
That days was also the day of the Penguin Dinner, which was at a walking distance from the University.
And I must confess I passed that dinner as a lzay guy speaking french with Thomas, and also 2 other french persons from Neo Caledonia ! We were near the Swan River and beloeve me, it was pretty cold compared to the Monday, so I ended up having a small flu :-( But that was a nice time remaking the world again and discussing technology.
The place was really neat with a view over the Swan River up to the city, and we stayed there a couple of hours, before going back to our hotels, tired, but happy of the day, and just waiting for the next one !
BTW all videos are now online so you can easily make your own opinion by looking yourself at them, as if you were there (you just lack the sun, the wine, and the nice talks ;-)) Cf: http://mirror.linux.org.au/linux.conf.au/2014/