MPU Meeting Monday 4th May 2009

Power Management Project

Chris has now worked out the correct suspend command/options for the Dell Optiplex 755. He also discovered that the patched intel graphics driver for SL5.3 does cope with the machine being suspended.

The machine gothenburg is still seeing occasional hangs though, he is trying a different machine to ensure that this is not related to that specific machine.

Chris has noticed that since SL5.3 there have been problems with machines not waking up at the required time, wake-up from the power button still works correctly though. He will try the latest test kernel with a stable machine using SL5.2 to see if the problem is kernel related. To use the latest test kernel the TEST_KERNEL and OPENAFS_TEST_KERNEL macros must be set at the top of the LCFG source profile.

rpmsubmit

Alastair has written a daemon which runs a script periodically to check buckets for changes and release volumes when needed.

It looks like there is an AFS (or perl-AFS) bug which means that signals do not always get returned.

There is also a new script to create repositories, this has been packaged.

Alastair is now looking at web serving the packages from AFS.

The rpmsubmit tool has been hacked to work with AFS. For now this is just a case of doing the minimum changes required, later in the project this will be revisited and done in a better way.

AFS Component

This is the status report for the development meeting.

Client support

The new component can configure the ThisCell file, the client cacheinfo file and the local CellServDB file. The only missing configuration step is to call 'fs newcell' to inform a running client of any new cells or changes to existing cells. That change requires new Perl XS code to be written for the AFS module.

The client arguments in the openafs sysconfig file can be manipulated via LCFG resources.

The component can also start and stop the openafs client. There is an option to not run the client which will be useful on the AFS servers. On starting the client the component can optionally turn on gcpags and encryption.

Server Support

The new component can manage the ThisCell and CellServDB configurations and the list of privileged users for the server. If the bos server is running then it will use the relevant bos commands otherwise it will generate the configuration files directly.

The bos arguments in the openafs sysconfig file can be manipulated via LCFG resources.

The component can also manage the individual service instances (e.g. buserver, fs, ptserver, vlserver). Currently it will create the instances if they do not already exist, it does not yet handle stopping, deleting and modifying running instances.

As with the client it can start and stop the openafs server and there is an option to control whether it is actually started.

Monitoring

A translator module is provided which uses the server resources to generate a nagios configuration for the openafs nagios plugins provided by Russ Allbery. These monitor bos, check disk space usage and look for blocked connections on file servers and also check that the Ubik replication between the database servers is running correctly.

TIBs

Chris has looked at this a bit to see what needs to be done, not actually started development yet though.

Miscellaneous Development

bugzilla
There is now a review flag, to allow users to request that other users review their patches. There was also a suggestion that we switch from bug owners to QA contacts for the dummy mailing list style watchers as people using this system are not seeing all the comments.

Operational

dresden data
We need to move the lcfg.org data which is mounted on dresden from the satablade to one of the new evo arrays, Chris will look into this.

SL5 kernel
The latest SL5 kernel has been built and deployed to the develop machines.

DIY DICE and roaming VMware
Alastair has gone back through the instructions and spotted that they were missing a crucial reboot step.

bugs.lcfg.org
Chris has finished tidying up the headers

Moving VMs
Alastair has documented how to migrate VMs between servers.

twiki upgrade
The twiki instance being used for wiki.lcfg.org needs to be upgraded. This is currently proving rather problematic, it appears a bug has been introduced somewhere.

lcfg-server update
Kenny is going to test this on his profiles and give feedback. If all goes well the plan is to ship this in the next testing release on 11/5/09.

live_testing_defaults.rpm
This is getting rather long again, we will raise the issue at the next Operational Meeting.

This Week

Alastair will:

  • Review mini-projects
  • Add sleep component to machines
  • Think about virtualisation mini-talk
  • Talk to Craig about dual path/controller
  • Work on rpmsubmit project

Chris will:

  • Review mini-projects
  • Chase up sleep problems
  • Test more hardware models
  • Look at TIBs
  • Finish dresden dual-path changes
  • Move dresden data to evo array

Stephen will:

  • Review mini-projects
  • Add sleep component to machines
  • AFS component
  • twiki upgrade

-- StephenQuinney - 05 May 2009

Topic revision: r1 - 05 May 2009 - 14:33:00 - StephenQuinney
 
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