MPU Meeting Tuesday 6th May 2014

Virtual DICE

Nothing happened.

LCFG Client Refactoring

Nothing happened.

LCFG systemd

Nothing happened, but Alastair is going to revive this project.

LCFG Port to RHEL7 or compatible

The current state of play can be seen at F20 components status.

  • Stephen's new grub2 component is as finished as it currently needs to be.
  • Several components have been rewritten into Perl.
  • We need to find out how to get kernel modules built at boot time in the systemd world. The kernel module needs to do this.
  • Stephen rewrote the hardware component, in the process removing some very obsolete features which appear to have fallen into disuse years ago. The new hardware component hasn't been installed on SL6 as it's a big change, but it might well be done when we bump SL6 up to 6.5.
  • Only one RPM can own a directory now. This will break RPMs still using the old tehnique of claiming ownership of a directory rather than listing every individual file in that directory supplied by the RPM.
  • Red Hat has made available a Release Candidate of Enterprise Linux 7. Both source and binary RPMs were freely available so we have mirrored them. No updates are being made available of course. Use of this release seems to be unlimited.
  • There is also a beta release of an EPEL 7 repository for the RHEL 7 beta release (see the EPEL page and the EPEL 7 beta FAQ for more details). We have a mirror of that too.
  • The LCFG Deployers' Meeting agreed that we should now move our development effort from Fedora 20 to the RHEL 7 RC release. Our short name for this, and for the final release when a free version comes out, will be el7. Our basic RHEL 7 release will be el7. Internally we will maintain Scientific Linux 7 (sl7) as a set of differences to el7. (Note that this is not a decision on which release we will use for the majority of DICE machines.)
  • Stephen has added EL7 support to PackageForge and has used it to rebuild our F20 packages for EL7. He will make the rest of the necessary platform infrastructure including the full range of package buckets. As soon as this has been done people in all units will be able to pitch in and start building and testing packages on EL7.
  • At the moment the install procedure is to install from an ISO (look under /afs/inf.ed.ac.uk/pkgs/sites/el7/7.0 or on KVM servers); then install lcfg-client using yum; use the client to download the LCFG profile; use yum to install updaterpms; then use that to install the rest of the local management tools. As yet there is no LCFG management of systemd (unless you use the file component) so components do not yet automatically start/stop/run appropriately, they have to be manually controlled or systemd has to be configured.
  • We'll need to discuss the details of starting/stopping/running components with systemd. We envisage providing some recommended/example systemd configuration recipes for components.
  • Chris will micro-tag and auto-build all remaining (Linux-supporting) LCFG components for el7.

Systemd

Alastair will now restart the systemd project and will make the systemd LCFG component usable on el7. The matter of how to use it in the el7 project is a matter for that project; the systemd project will just deliver a working component and then be wound up.

Miscellaneous Development

  • Stephen has rewritten the way logrotate templates are handled in ngeneric. This was prompted by the discovery that if the same file is listed more than once in the logrotate configuration, logrotate simply crashes. Existing logrotate configurations will continue to work as before. In addition it's now possible to use Perl Template Toolkit for a logrotate template. The extralogs resource has been extended to make it possible to specify an absolute file location by starting it with a slash. It's now also possible to replace just one block of the standard logrotate configuration rather than the whole thing. This means that subsequent changes in the rest of the standard configuration will be inherited.
  • Graham kindly provided Chris with some patches to extend the diskfull script. It now optionally sends output to the shell and indicates status using its return value.

Operational

  • Chris now has enough information on our HP servers that it seems likely that we would be able to download even firmware updates which were restricted to HP customers with support contracts.
  • Alastair has ordered our NEW LCFG master and package master servers. They will be Dell R210 II with a RAID pair of 10k 300GB SAS disks. Each will have 16GB of memory and a five year warranty.
  • Alastair has done some latency testing on a KVM server. The conclusion is that to avoid performance problems our KVM servers shouldn't have more than four VMs per physical core.
  • Following this work Alastair will check that the pinning of the LCFG slaves is still appropriate.
  • Chris will check the config of staff.nx then announce it.

Next Meeting

Wednesday 21 May at 2pm.

This Week

  • Alastair
    • Order a spare 600GB disk for waterloo (hot spare)
    • Double check latest web security reports
    • systemd project
      • start writing in blog
      • Modify lcfg components/rc scripts list as a result of COs talk.
      • Complete lcfg-systemd component
      • Update SystemdSchemaProposal
      • Consider how components will work with systemd
      • Consider journald
    • Order new kit
    • Add more memory to Forum KVM servers? - 700 per server to upgrade 64GB -> 128GB
    • Check pinning of LCFG slaves still correct (wrt latency testing) Slaves pinned to one physical CPU

  • Chris
    • SL7
      • Continue looking at systemd
      • Look at autobuilding components
    • open up staff.nx and announce (check identical to existing nx service)

  • Stephen
    • LCFG client refactor stage 1 -> activity page
      • schedule debrief meeting
    • LCFG client refactor stage 2 -> activity page
      • continue development and docs
    • Check with SEE what they did to improve NX performance -> activity page
      • make any easy changes
    • SL7
      • Switch to EL7
      • ed/dice flavours of inf level
    • Reboot hare to test firmware update
    • Pandemic stuff
      • discuss school db with Graham/Tim
Topic revision: r6 - 20 May 2014 - 09:31:36 - 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