MPU Meeting Tuesday 23rd October 2007

LCFG Website

It's been proposed for closure at the next development meeting.

Solaris Improvement Project

Nothing happened.

SL5 Project

It's been proposed for closure at the next development meeting.

Chris was worried about possible requests for documentation on inf-level SL5. There shouldn't really be any need for this as the inf level isn't intended to be used, but one solution might be to initiate the SL5 Upgrade page with some brief details of the current state of how to get an inf-level machine up and running.

Buildtools Project

Stephen's going to talk to Kenny this week to get his views.

rpmsubmit Project

No changes this week.

Paul suggested an updaterpms option which submits to the same bucket as the previous version was submitted to. This is outside the scope of the current updaterpms project which is solely a reimplementation of the repositories using AFS instead of NFS.

Power Management

Chris asked if anyone had ideas on where to start, apart from wikipedia and lesswatts.org, and got plenty, so thank you very much: suspend technology; Tux On Ice; powerd; Fedora release notes; the proceedings of the Linux Symposium, a conference held in Ottawa every year.

LCFG future

We're looking at funding for a rewrite of the LCFG core.

Also, there's a possible SAGE LCFG book in the works. Paul will produce an outline of its content and will ask for comments.

LCFG Deployers Meeting

Kernel modules
A pre-meeting confab with Kenny produced a two-part solution to the kernel modules problem. Firstly, Stephen has made two new package lists, lcfg_sl5_kernel.rpms and lcfg_sl5_64_kernel.rpms. These are for kernel module packages for SL5. Kernel module packages for SL5 will only appear in these files. They will thus be excluded from the normal updates package lists. Secondly, Alastair is changing updaterpms such that the "only on reboot" package spec flag will apply to package removals as well as package installations. The kernel package lists are up and running (so SL5 installs now work again) and the new updaterpms will be packaged, tested and installed in the next week or two. Until the new updaterpms is deployed, existing SL5 machines should have updaterpms -f run before any rebooting to prevent them being upgraded into an unusable state.

How to handle SL5 updates
The meeting considered how to react when SL goes from 5.0 to 5.1, or 5.2. These will be released in the form of updates. We're going to deploy major SL5 updates via the test_updates package list, testing the updates over at least a couple of release cycles. Unfortunately this leads to other problems: where do we then put the updated packages - in the base list? In an ever-expanding updates list? What do we do with the previous base list? How long can we keep its packages? How many versions of the base software do we support at once? And how long for? What do we mean by support in this context? We decided to postpone consideration of these questions. One thing that seems fairly clear is the need to redesign the header structure to encompass the possibility of support for multiple releases, maybe with different releases being supported by different organisations.

Operational

LCFG Zero directory bug
Kenny found a bug: the file component couldn't make a directory called 0 (zero). This turned out to be because of a problem with how our Perl code tests values. The bug crops up in multiple places. Stephen has found and fixed instances of it in lcfg-ngeneric and lcfg-utils.

Berkeley DB
The zero directory bug reminded Alastair of another bug which is worth noting: when using Berkeley DB in C, if you don't allocate the correct resources first, you can run out of resources, triggering a hang which can persist even after a reboot (maybe because there's a cache involved somewhere). Perhaps this could be something to do with the periodic updaterpms hangs?

Autoreboot bug
Once autoreboot has triggered a shutdown, you can't initiate a shutdown by any other means, as only one shutdown can happen at once. Oops. It needs to be fixed and we think it should be pulled from this week's release and fixed.

Newsletter
Here are some likely topics for the next edition: SL5 on servers; 64 bit support; the possibility of SL5 on desktops next summer.

iFriend for LCFG wiki
Paul asked if the LCFG wiki could use iFriend. It doesn't yet but yes it could, and this should happen as part of the "apacheconf for dresden" action. The action's description has been expanded to make this more clear.

Tests failing
Various lcfg-server regression tests were failing because of minor changes to various things. Paul has tidied them up, and they now succeed again, but he wants to emphasise the importance of keeping the tests working.

This Week

Alastair will:

  • RT Rota
  • Package and test new updaterpms.
  • More on rpmsubmit
  • Sift upgrade wikis for wisdom
  • Speak to George about Mac headers

Chris will:

  • Start investigating power management.
  • Solaris patching
  • Look more at the twiki bug plugin.
  • Document inf-level SL5.

Stephen will:

  • Virtualisation write-up
  • Start buildtools project
  • Reboot tobermory

-- ChrisCooke - 24 Oct 2007

Topic revision: r2 - 24 Oct 2007 - 13:22:35 - ChrisCooke
 
This site is powered by the TWiki collaboration platformCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback
This Wiki uses Cookies