TWiki> DICE Web>MPUnitMeeting20090513 (revision 2)EditAttach

MPU Meeting Wednesday 13th May 2009

Power Management Project

Alastair has added the sleep component to one of his machines, a T3400, and is working on finding a suitable display quirk to use in the suspend command.

The occasional (every 50-100 sleeps?) machine hangs seem to be related to the new kernel: they happen on all 5.3 machines Chris has tested, and also on a 5.2 machine using the latest test kernel, but not on a 5.2 machine with a standard kernel.

The failure to wake up has not been observed again.

Tim tested the component on his desktop machine and it uncovered a gap in the idleness checking. The component assumes that the machine is idle if the load average is below a certain level (default 0.1) and all interactive shells have been idle for more than a certain time (default 10 minutes). However some activities can be compatible with the machine fulfilling both conditions, and Tim found his machine going to sleep as he was editing a long document. Chris has therefore had a rethink and (to cut a long story short) now reckons that we should let X session managers decide for themselves when the machine is idle. In Gnome, gnome-power-manager does this perfectly well, and can be configured to suspend the machine a number of minutes after the machine becomes idle. We can use gconf and James' lcfg-gconf to set this as default behaviour for all gnome users. (User preferences can override default behaviour. If necessary gconf can also set mandatory settings which cannot be overridden by user preferences.) This takes care of idleness detection and suspend; resume will still be handled by the sleep component, which will continue to run every minute or two and will ensure that the machine always has a suitable wake time set. The machine will therefore not miss any important cron jobs no matter what puts it to sleep. (Coping with user-initiated suspend was always part of the sleep component design anyway.)

Chris will now concentrate on fixing the idleness/interactivity detection problem above and not worry too much for now about occasional hangs; and get the component into the develop release.


The daemon now supports nagios monitoring - so for example a configuration fault now results in a nagios warning.

The AFS signal problem was caused by calling vos now in every bucket. This should have worked but didn't; it's been changed; things are better now.

Some work has been done on serving packages over the web: Alastair has got waklog renewing AFS credentials successfully.

Alastair has done some benchmarks and found that repositories on virtual machines are 6-7 times slower to submit packages to than those on real machines, for both old and new repository types.

Most documentation has been done, the exception being the architecture document.

AFS Component

It's mostly done. Some code is still needed for the perl AFS module for client support. The component itself has been documented, but descriptions of how to use it are still needed.

The component's being used to configure and run two database servers and a fileserver. The database servers successfully communicate with each other. A test fileserver will be used as a guinea pig. Database migration should be straightforward - just set up extra virtual database servers alongside the current real ones and gradually migrate to them.

64 bit client support still needs to be sorted out.


Craig's working on a requirements doc for the component.

LCFG Core Refactoring

A project meeting was held and a plan agreed - see Alastair will update the brainstorm writeup.

Miscellaneous Development

We reviewed the mini projects list.

The profile for prague is at last clean and tidy - the forumtracker setup is fully in headers.

Alastair has found out that it's not too late to work on dual path/controller improvements. Craig has promised to move hawthorn to KB so that it can participate in experiments.


The central disk blob will move on Thursday.

Stephen mentioned live_testing_defaults.rpms at the Operational meeting. We'll look at it in a month to see how much it's improved.

Roger's upgraded twiki seems to work happily with iFriend, so Stephen will try copying Roger's config for the lcfg wiki upgrade.

Stephen will fix the cmirror LCFG level broken dependencies.

The Perl component code ignored quiet status - Stephen has fixed it.

The AFS cachesize value seems to be too small at 1GB - the cache partition is 2GB. Stephen will test a bigger value. The maximum size shouldn't be more than 90% of the available space.

Alastair will update the colocation plan. We should move mousa from FH to KB at some point.

When accounting for time management could we please all break down our operational work for a while into smaller categories? This may suggest new permanent categories.

We can forget the FC5 release testing, it's now a waste of Richard's time. Stephen will tell Richard.

Carol could make the testing release each week. Chris will arrange it with her.

This Week

Alastair will:

  • Tidy the mini-projects list
  • update the colocation plan
  • Talk to Carol about the LCFG level work.
  • Think about a virtualisation mini-talk.
  • Remind Craig about moving hawthorn to KB.
  • Work on rpmsubmit project.

Carol will:

  • Set up LCFG/inf level VM to monitor LCFG level.
  • Learn how to make the testing release.

Chris will:

  • Hand over the making of testing releases to Carol.
  • Finish fixing the sleep idleness detection problem.
  • Move dresden data to evo array.

Stephen will:

  • fix cmirror dependencies.
  • test a bigger afs cachesize.
  • Let Richard know that fc5 release testing isn't needed any more.
  • Test the sleep component.
  • Finish the AFS component.
  • Upgrade lcfg twiki.
  • Tackle 'om' improvements.

-- ChrisCooke - 14 May 2009

Edit | Attach | Print version | History: r4 < r3 < r2 < r1 | Backlinks | Raw View | Raw edit | More topic actions...
Topic revision: r2 - 15 May 2009 - 13:12:59 - 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