MPU Meeting Tuesday 21st August 2012

Simple KVM Service

(FX: Fanfare) The project has gone for signoff. KVM developments will be tracked in the Miscellaneous Developments section from now on.

Server SL6 Upgrades

Stephen has split the bugzilla packages into SL and EPEL categories.

Chris has made some progress with the dependencies for the Bugzilla 4.2.2 RPM from A key breakthrough was to translate the automatic dependency filtering from RPM 4.9 format to RPM 4.8 format. This page explains both the old and the new types of dependency filter macros.

Server Hardware

The code has been checked against all server models we have and should now discover most hardware components' details correctly most of the time. The code to insert the results into the database is ready to deploy. Coming up: add a report which compares the discovered firmware versions with known good recent versions.

Security Enhancements

The data retrieval script is now running every day and seems to be working smoothly. Report generation is almost there too - it just needs a few minor tweaks before being ready. The documenting of the report generation isn't yet complete.

Stephen suggested that monthly reports would be useful in addition to daily reports, as they would highlight machines having frequent intermittent problems.

Stephen has talked to Alison and Carol, who will compose some messages to be sent to people whose machines have suffered (for example) kernel oops and oom killer events.

Miscellaneous Development

A further boot component fix
The boot component had previously been fixed so that when launching an rc script it closed the 11 and 12 file descriptors. It has now been changed so that the file descriptors are also closed when a component is launched using om. This issue was causing big problems for auditd as well as quite visible ones for anything launched from the cron component, for example the KVM servers' LVM report script.

New kvmtool
Alastair has shipped the new kvmtool. It can now discover a guest's host server automatically, and there are new console, edit and start commands. Still to come: shutdown and destroy commands. Stephen suggested a simple info command and Alastair will add that too.

KVM cold migration
Alastair has documented this. See Cold migrating a guest from one server to another and Cold migrating a guest from one volume group to another (on same host). He will next document the process of moving a Fibre Channel based LVM volume group from one KVM server to another.

KVM console logs
Alastair has investigated this. It's non-trivial even with conserver. A lot of people want this. There's a summary of the issues at Stephen floated the idea of having multiple console devices at the operating system level, some of which could be logged and some interacted with. Alastair will follow up this idea.

Installroot additions
Following the recent dunlin incident, Alastair and Stephen have added some extra utilities to the installroot and have ensured that it can now mount USB keys. While doing this Alastair noticed some legacy stuff which now needs to be removed, for example rc_install.

SL 6.3
Stephen has been working on the SL 6.3 upgrade. The lcfg level is now done. The dice level is mostly done. Machine profiles are mostly fine, except for four which are beyond the pale. He still needs to fix the profiles used for PXE and the install CDs. We should soon have some real DICE machine profiles using SL 6.3. You can try 6.3 by defining DICE_TRY_SL63 at the top of a machine's LCFG file. We'll put the timing of the 6.3 upgrade to the Operational Meeting.


PXE and LCFG Master moves
Stephen has moved PXE from schiff to hare. He's also removed schiff from cache.pkgs. This leaves schiff ready to be the next LCFG Master, taking over from tobermory. Chris will tackle the LCFG Master move.

Bakerloo reboot
Stephen rebooted bakerloo and it's worth noting that its KVM guests suspended and resumed successfully with no problems.

Server reboots
we will go ahead and reboot northern, piccadilly and circle as they are mostly running less important guests. The others will be done later.

7900 sleep
Chris left sleep enabled on a couple of machines before going on holiday and found no problems on his return, so he will reenable sleep on 7900s this week.

DBAN Quick Wipe
We need an additional PXE option to do a quicker DBAN disk wipe than the existing DBAN option. A balance should be struck between speed and effectiveness. Chris will do this.

Chris will fix its apache config.

2012-2015 MPU Hardware Upgrades
Chris will bring the ServerBuying2011to14 page up to date with recent server deployments.

This Week

  • Alastair
    • Fix permissions problems on ordershost - awaiting Graham's return to explain things
    • kvmtool - destroy and any other simple commands + a search + info display of guest (eg cpus, memory, bootorder, how many disks etc)
    • document moving fibre based volume groups from server to server Done as much can be done without testing
    • Look at whether can get kernel to log to two devices, and capture output of 2nd device to file ?This doesn't work. But, it looks like conserver might.
    • Pull out rc_install and other legacy stuff out of lcfg-install
    • Work through LCFG bugs
    • Meet with Stephen
    • Code review Roger's package search tool -
    • Personal development topics
    • Server replacement plan for 2012-2015

  • Chris
    • Complete hardware monitor data -> orders database. Propose a manual process for using this information
    • Bugzilla backport
    • Create short wipe dban pxe option
    • Fix dr.pkgs apache config so it works
    • Prune activities list
    • Personal development topics
    • Server replacement plan for 2012-2015

  • Stephen
    • Continue with SL6.3
    • Security project
    • Organise openafs conference
    • Server replacement plan for 2012-2015
    • Personal development topics
    • Meet with Alastair

-- AlastairScobie - 21 Aug 2012

Topic revision: r9 - 27 Aug 2012 - 15:11:47 - 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