MPU Meeting Tuesday 14th February 2012

LCFG Server Refactoring

A minor problem has come to light with the new mutation code. It happens when undefined values are used in mutations. The undefined values do correctly end up as empty strings in the final mutated resource value, but warnings are generated. We view undefined values as acceptable in mutations and assume that they should end up as empty strings, so the warnings are not needed. However it's going to take some careful thought to correctly eliminate the warnings. This is LCFG Bug 533.

The server log on mousa is still reporting periodic rsync timeout errors. Chris will try increasing the timeout to two seconds to see if that helps.

Wake On LAN

The project was signed off at the 1st February development meeting with this final report.

The wake.inf.ed.ac.uk service has been moved from adriatic (VMware-based) to ionian (KVM-based on bakerloo).

Simple KVM Service

rvirsh has not yet been made available. Alastair will chase Iain.

SL6 Server Upgrades

Chris has been building up an SL6 version of tobermory, the LCFG master, on a KVM instance called tasman. So far subversion seems to be working properly, but there are problems with websvn.

Chris will mail COs about the defaults packages missing on SL6.

Chris has upgraded mousa to SL6.

Miscellaneous Development

Autoreboot
LCFG Bug 532 - the component was logging all standard error output from at as errors. Stephen fixed this by making the component log such output at the Info level unless the command actually failed. Fixed in version 1.0.14 of the autoreboot component, in the stable release on 2012/02/15.
Grub
Stephen reported another fix to get the new uuid code working properly with installs. The code wasn't getting the right UUIDs for devices at install time. Alastair suggested using blkid (which gives a uuid when given a device name) and when Stephen tried this it solved the problem. He also changed it to use /proc/mounts instead of /etc/mtab. Once this week's stable release is out, the installroot should be updated; then the new uuid-related grub resources should be introduced; then we can switch to booting using uuids.
dhcpd
Stephen added a uid and gid for dhcpd to the defetc package for Ian's SL6 server upgrade.
Multipath
Alastair has merged the mpath and multipath components.

Operational

  • A machine called patton was sending lots of rootmail related to the sleep component. Chris found that this was happening because the sleep component was being uninstalled when the machine was switched into exam mode; which meant that when the machine was switched out of exam mode, the sleep component was being started and repeatedly called before it was reinstalled on the machine. The component should not be uninstalled during exam mode; Chris will fix this.
  • Alastair and Stephen have been investigating a problem with the boot component. In some circumstances there seems to be a problem whereby the boot component detects what it thinks is a stale timestamp on the boot component's run file, causing it to delete the run file. From that point onwards the boot component appears not to have been started. To help debug this more fully, Stephen is adding debug code to the ngeneric component to make it print out details of the boot component's status files.
  • Alastair tried Wake On LAN with the ATI card removed from his HP dc7900. Removing the card made no difference.
  • Alastair has asked Alison to sort out the spare memory.
  • Chris has moved openldap-bdb-locks-increase from individual LCFG server profiles to the LCFG slave and master headers.
  • Chris has put local home directories on most MPU servers.
  • Stephen raised the creation of lcfg/options/virtual-hardware.h at the Deployers Meeting. The idea was favourably received so Chris will create the header.
  • Stephen has asked COs to use bakerloo for important VMs from now on (rather than northern or circle).
  • Chris has moved wake.inf.ed.ac.uk to bakerloo - specifically to a VM called ionian.

This Week

  • Alastair
    • Chase Iain about bashdefenv
    • Arrange figgy (with replacement RAID card) to go to KB
    • Help Chris with SL6 inf layer (for ashkenazy)
    • Discuss with George - how rack desktop servers in AT and IF
    • Learn more about Theon

  • Chris
    • SL6 defaults list
    • Continue upgrade of tobermory to SL6
    • SL6 inf layer
    • Replace ashkenazy with new KVM based SL6 server (on northern) - and compare profile generated with that generated by ashkenazy
    • PD - local url shortener implementation (for not-a-service)

  • Stephen
    • Update the PXE installroot
    • AFS talk
    • Theon
    • Consider PD - concrete task

-- AlastairScobie - 14 Feb 2012

Topic revision: r3 - 17 Feb 2012 - 12:20:08 - 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