MPU Meeting Thursday 23rd July 2015

LCFG Client Refactoring

Nothing happened

Inventory

Chris and Alastair met with Graham last Friday to discuss the theon conduits which will needed for the new inventory, Alastair will write up the notes. The simplest approach might be to just connect directly to the school DB and do SQL queries, another alternative is to use CVS files as a transport format. The discussions revealed a problem with querying the room information, it should be possible to resolve the issues with the creation of some views.

Alastair has improved the current data some more. In particular the handling of items which are "resold" from the school onto individual grants has been improved.

Whilst looking at the machine location information Alastair has realised it's all broken because of the moves to FH and Wilkie. The Inf Unit needs to commit to an API which can be relied upon to not change (without prior warning).

Miscellaneous Development

rdisc problems
We have had to deal with rdisc not working on freshly installed SL7.1 machines. The investigations revealed a number of issues. (1) The initramfs file was not being rebuilt when systemd or dracut changed, this was particularly an issue since the SL7.1 upgrade did not include a kernel upgrade - which would have provoked a rebuild of the initramfs. (2) The kernel component needs to use the new-kernel-pkg script to rebuild the initramfs rather than calling dracut directly. (3) rdisc does not play nicely with networkmanager. Thankfully the new version of networkmanager has a configure-and-quit option which can be enabled. This means that once an interface has been configured it is not touched by networkmanager again, this allows rdisc to then manage the routing. (4) rdisc does not properly handle ioctl failures, it just prints an error message which can leave things in a bad state from which it cannot recover.

auditd
Stephen has removed the auditd component from the SL7 installbase. This guarantees that installs are much quicker. There was some discussion about which contexts are applicable, it appears that the installbase context can be used in the 2nd stage of the install.

mirror-scripts
Stephen has finished the new mirror scripts, this includes the various patches from Kenny. There is now full documentation of how to use and configure the scripts and the MPUOsUpdates page has been updated. We still need to unmount the old sites pkgs tree in AFS and decommission telford.

Operational

New KVM server
The new KVM server is named azul. It has been setup without the br0 interface.

waterloo
The KVM server waterloo has been updated to SL6.6 and the memory has been doubled. It is going to need another reboot to swap the new partition to /var/lib/libvirt and to sort out the network bonding.

Server moves
Chris has wiped district and moved it to the junk heap. He has also moved atom out of the way to clear rack space.

New CA certificate
Stephen will distribute a list of MPU servers which need new certificates. We also need to update the eucs-sslcerts package, firstly to add the new certificate and then to remove it entirely.

SL6.6 reboots
Stephen will reboot the SSH servers and the sites package server (juice). Alastair will reboot the package caches and an NX server (northern). Chris will do the packages master (bruegel) and the other NX server (piccadilly).

7.1 rollout
We should aim to roll out the 7.1 upgrade with the stable release on 5th August.

This Week

  • Alastair
    • Inventory project
      • continue working through TartarusWorkFlow
      • finish deploying new order file processing code
      • finish document new parser and API
      • consider what next can be integrated into existing system, if anything
      • write up meeting with Graham re conduits Awaiting Chris and Graham's comments
    • @home - look at using rsync from site.pkgs instead of mirroring from upstream
    • Need to remove default bridge from kvmtool create
      • make necessary changes to live/mpu_kvm_server.h (support both methods)
      • Look at wrt azul.Chris has done. ??? Do we still want to remove the default bridge ??
    • Remove default pool if ops meeting agrees
    • Schedule firmware upgrade for DS3254
    • Take creation of mdp-installbase profiles to LCFG deployer's meeting (bug #878) (On July meeting agenda)
    • Add instructions on adding extra disks to an LVM/KVM pool - to KVM host guide
    • Talk to Phil Wadler about machine upgrade - ask for a list of software he requires built
    • Experiment with different window managers under VNC (making the assumption that performance under NX will be similar)
    • Look at pkgsearch code (to see vulnerability)
    • Schedule reboot of oyster with new memory install - Expand suspend space- needs a reboot to pick up /dev/sda4 - move /var/libvirt before KVM clients come up -await new memory Check using eth0 and eth2
    • Think of a use for 'atom'
    • Convert kvmtool to support ini file as well as yaml config Still to ship
    • Ask Graham want atlas package is user_common for - is this a real problem? It isn't installed on servers
    • Understand how NetworkManager works wrt init scripts
    • Create a page documenting how to find packages (simple version for now - locate on salamanca)
    • Order up 16 of 8GB DIMMs for oyster and hammersmith (ASAP)
    • Reboot northern (schedule ASAP)
    • Reboot hare and wildcat

  • Chris
    • Inventory project
    • Finish work on SL 7.1
      • test installer after stable release of 15/07
    • pkgsearch for SL7
    • Expand suspend space on waterloo
    • Coordinate move of district to junk rack and harvest disks
    • Look at inifile component configuration of kvmtool
    • Start looking at SL7 Virtual DICE
    • Try out updated MPUOsUpdates instructions
    • Reboot piccadilly and bruegel

  • Stephen
    • LCFG client refactor stage 1
      • schedule debrief meeting
    • LCFG client refactor stage 2
      • finish off C code
    • Think about PD - Interested in ZeroMQ
    • Finish off window manager selector code
    • Circulate a list of MPU servers to reboot for SL6.6
    • Fix kernel component to use /sbin/new-kernel-pkg (On sl7.1) to rebuilt initramfs
    • Check whether installbase context is working as advertised - https://wiki.lcfg.org/bin/view/LCFG/InstallContexts
    • Produce a list of MPU servers which don't use Terena certs
    • Reboot brendel and schiff and juice

-- AlastairScobie - 23 Jul 2015

Topic revision: r10 - 23 Sep 2019 - 13:33:37 - 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