MPU Meeting Tuesday 27th July 2010

Project Planning

The next project planning round is almost upon us. We need to provide figures and proposed projects by 2 August, although in practice some of the detail can be postponed until 9 August. We shall meet on Thursday at 3pm to thrash this out.

LCFG Server Refactoring

Rescuing git repositories
The git repository for this project is on the not-a-service machine which is not in a good state. Stephen will move not-a-service to blurt to enable work on the project to continue.

Stephen is going to start the process of testing the changes. It takes a few hours to do each test so this will take a while. Once the tests are OK it will be possible to release an updated version of the server software soon afterwards.

Server Hardware

This project ran out of time and seems to have settled down so it's suggested that Chris put it forward for sign-off next week at the development meeting. We should discuss possible follow-on work at this Thursday's project planning meeting. We should record the possibilities for follow-on work. Chris will provide a final report for the project.


Alastair has got the new installer working with SL5.5 and fixed a bug he had introduced. We won't use the new installer with 5.5 but we will with 5.6 when that comes along.

We currently need a running F12 machine to make the installroot for F13. Alastair will work to fix this situation. Alastair expects to be able to put the project forward for signoff at the September meeting.


Make xauth use /tmp mnot homedir
Stephen will look at this.

Package Lists
All the LCFG-level package lists have now been sorted out and conflicts, dependencies and so on eliminated.

Testing of Package Lists
Stephen will patch up his automatic testing of various package list combinations such that it can form part of the weekly release testing. As it will take hours to run all of the tests he envisages it being done overnight by cron.

F13 desktop list
It's included differently now: when the DESKTOP_PACKAGES macro is set, the lcfg f13 base list includes the lcfg f13 desktop list at the bottom. This seems to work.

F13 lab upgrade script
Stephen has written one. It will change the profiles of all machines which the inventory thinks are in a particular room.

Outdated lab machine profiles
Three hundred machines are using student lab headers in their profiles but we only physically have room for two hundred machines at the very most! We suspect that many decomissioned machines have not had their profiles changed appropriately. Student helpers will sort this out.

The Inventory
We really need to sort it out!

Bash defenv
This is missing. Iain is working on it and it should be there soon.

Console issues
Alastair will fix these - for instance the X session hops about between VTs 7, 8 and 9 rather than being on VT 1.

GDM starts at the wrong time
GDM starts at inappropriate times such as when reboot is imminent.

X login is very slow
Is this because it runs lots and lots of helper apps? We should review the list of things which start up on login.

Nautilus creates directories in your top level homedir
but then again so does everything else these days so we'll either have to put up with it or use very old software.

Partitioning problem
Alastair needs to fix this.

LCFG port to F12

Chris will write a final report and put this forward for sign-off in August. Alastair has a figure that may help: we spent 14.3 weeks on this all told, when we expected to spend 8 weeks on it.

Miscellaneous Development

Panos' changes for the network component
Stephen will review them.

Resolvconf in DICE installbase
The current DICE installbase works so Stephen sees no benefit in interfering with it. Alastair will investigate and see if there is a reason it needs to be done.

SL 5.5
This will go out in this week's stable release.

PXE change to 5.5
After this week's stable release Stephen will update PXE to use SL 5.5 images.

New yum component
Stephen has written a yum component to manage yum repository configurations. This will simplify the setup of yum on our machines. We still use updaterpms for software updates but yum and yumdownloader are both useful when searching for and deploying software.

Updated mock component
Stephen has updated the mock component to make it always run the createrepo script. This creates a yum repository. This can be really useful when a large number of packages are being built as packages can be added to the repository as they are built and then be available as dependencies for subsequent package builds.

qxpack fix
Stephen has fixed qxpack to stop it making silly guesses about architecture.

Personal Development Topics

We should all think about this topic this week and come up with some personal development plans.


Big updates on virtual servers
When SL5.5 hits stable on Thursday we want to avoid the problems we had the last time every virtual server tried to apply a large update at the same time. The stop-gap solution will be to define DICE_STICK_WITH_SL54 on virtual servers, then selectively remove it from a few servers at a time over several nights until all the updates have been done. Alastair will take care of this.

Routing delay problem
Alastair has spoken to George about this, but has since realised that the problem has been affecting us for more than a year now so will go back and speak to him again. The main symptom is big updaterpms delays on installs, delays which affect virtual machines particularly badly.

Metropolitan FC problem
Chris had looked at this and found that the FC cables were connected to different switches so failed to see a problem. However it seems that all FC switches on the same rack are closely interconnected and should be regarded as a single switch, so for better dual pathing the FC connections should go to different racks. Chris will adjust the cabling for metropolitan accordingly.

Unusual graphics card and monitor configurations
With the advent of F13 it is at last straightforward to set these up: you just need to use a configuration tool. We'll propose to the User Support Unit that one of their people documents how to go about using this tool to achieve various unusual setups.

New storage arrays
Alastair reports that they're overdue, they should have arrived last week.

Perl AFS module
Stephen has been working with the author of the module to fix its problems in our environment. The author is working towards a 2.6.3 release. Once we have a copy that passes its tests in our environment, we may need to fix up our scripts slightly as the API may change slightly in the new release.

Criticality information
Declaring and tracking criticality information properly would really be a mini-project, possibly involving a new mutation macro and a database. For now we should simply declare the resource in machine profiles and set up a weekly mailing listing critical machines.

This Week

Alastair will:

  • Spread out the SL5.5 updates on the virtual servers.
  • Talk to George about the routing problem.
  • Dual path _bpbeast_
  • Check DNS in the installroot and installbase.

Chris will:

  • Write reports.
  • Fix the metropolitan FC problem.
  • Work on Hadoop and other RAT stuff.

Stephen will:

  • Work on LCFG Server Refactoring.
  • Update PXE after the stable release.

-- ChrisCooke - 27 Jul 2010

Topic revision: r3 - 03 Aug 2010 - 13:47:31 - AlastairScobie
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