MPU Meeting Tuesday 23rd November 2010

LCFG Server Refactoring

Stalled.

Software Build Farm

Lots of small tweaks to allow the various daemons and tools run as a non-root user. Also added were the various kerberos principals and AFS users and groups for the various daemons.

Lots of time spent on ensuring that the daemons are able to shutdown gracefully and not leave the system in an inconsistent state. The incoming queue processor has also had lots of attention and is now mostly finished.

Tracking of the job status has now been added. During the incoming queue processing this shows which stage has been reached. After acceptance this is based on the aggregate of the statuses from the various tasks. The plan is that whenever a task status is updated the database will automatically recalculate the job status (not yet implemented).

Replacement for VMware Server

Alastair has set up a VM, named trailblazer, on the IS VMWare service. The install all worked fine with our PXE service. Alastair reckons the performance is similar to what we see on our server with the new storage array. The VMs are managed via vSphere, we would need access to a Windows Server to allow multiple RDP connections for different consoles. The chances are that IS will provide this as part of the service since lots of CSE are primarily Linux based.

To support KVM Alastair has reviewed Panos's patch for the LCFG network component which adds support for bridging. He has reworked it slightly and checked that it works with bonding, still need to check that it works with VLANs.

Alastair had to create an lcfg/options/newnewscsiroot.h header to add support for the vda partitions used by KVM. Stephen will review this change and merge it along with the lcfg/options/newscsiroot.h header into the lcfg/options/scsiroot.h header to simplify things.

Alastair also looked at the LCFG libvirt component written by Panos. It has been written in bash which is not great for something so complicated, it would be better in Perl. There are also lots of hardwired assumptions in the code. It looks like we would need to create a new component (or new code anyway) to manage KVM.

Chris and Alastair need to meet this week and work out who will be doing what in the project.

Miscellaneous Development

Date bug in build tools

Kenny spotted that the LCFG build tools have been using an American-style date format since they were created. Stephen quickly fixed it so that it works in a more sane way.

installroot

Alastair has applied a patch from Kenny which fixes bug#65 and bug#321. The initramfs was also updated for f13_64 to fix bug#351.

Operational

Optiplex 740

We have one Dell Optiplex 740 with an nvidia graphics card, named ord, which was being used in the 5.02 meeting room to drive the big TV screen. For some reason this has stopped working correctly with SL5 but it works fine with F13. We have temporarily switched to using an HP DC7900 in the meeting room whilst we investigate whether we can use F13 in the meeting rooms.

sauce

Our HP server, sauce, has been moved to KB and is now up and running. We now need to look into syncing package repositories on the local disk and configuring the machine so it can be used as a package server in emergencies. We also want to look into having it work as a backup LCFG master and slave for emergencies. This all needs to be reliable and simple so that any CO can restore service if necessary.

MPU VMs

We need to move some MPU VMs back to metropolitan now it has been fixed. It was decided that it would be easiest to just create new VMs for the external package server (currently genoa) and the DIY DICE server (currently padua) and then transfer the CNAMEs. Once they are moved we can attach bakerloo to the new storage array.

northern & piccadilly

These now have IPMI configured.

R710 and PXE

It was noted in the wiki that PXE does not work correctly with the Dell R710 and the work-around is to use the pxeclient.default_label resource.

This Week

• Alastair
  • Add atom to MPU server list
  • Meet with Chris to discuss virtualisation project
  • Think about package mirroring for sauce - add an rsync module to genoa's configuration to allow sauce access, and then use rmirror component to copy across
  • Virtualisation project
  • Look at wee projects list

• Chris
  • Meet with Alastair to discuss virtualisation project
  • Make trailblazer an LCFG server
  • Move genoa from bakerloo to metropolitan (prob reinstall)
  • Move padua from bakerloo to metropolitan (prob reinstall)
  • Virtualisation project
  • RAT work (submit)
  • Look at wee projects list
  • Move VMs from district to metropolitan
  • VMware Tools for f13 and f13_64
  • Make an f13_64 buildhost

• Gordon
  • Look at bugs and produce list of proposed kills and merges.
  • Look at wee projects list

• Stephen
  • Look at merging scsiroot, newscsiroot and newnewscsiroot
  • Look at LCFG slave/master on sauce
  • Move circle to AT
  • Sort out AFS access for genoa replacement
  • Buildfarm project
  • Look at wee projects list

-- AlastairScobie - 23 Nov 2010