Operational Meeting

Apologies for absence

Chris and Toby had sent their apologies.

minutes of the last meeting.

Craig pointed out that the reference to 4x1.5TB of SAN space in the Services Unit report should have been 4x1.6TB.

Report from Computing Executive Group

Alastair asked for any final comments on the proposed policy for platform support to be sent to him very quickly so that he could publicise the proposed policy more widely.

Ian Durkacz will be starting on Monday. His office will be JCMB 2417 (Neil's old office). Some concern was expressed that the HR feed to IDMS was going to be down from 16:30 today and not back till Monday (in case Ian's UUN was not yet determined). Carol offered to check that the UUN was set (and it was, to idurkacz).

Alastair has had further talks with the Head of School yesterday on the issue of computing staff accommodation and it looks as if his position on seven staff per 35 sq.m. room has softened slightly. Mike will be calling a meeting of all computing staff at some time in the near future.

Reports from units.

• Infrastructure.

  George reported that he had upgraded the last of the routers, solti, to FC5 but that he had been forced to turn off AFS on the machine because of a problem that Neil had precipitated. That just leaves most of the console servers to upgrade and basilisk to replace.

  At a recent meeting the Infrastructure Unit had decided to split the KDC master from the LDAP master (both currently on basilisk). This would make it easier to configure the disks in a way most appropriate for the two services, as their needs were dissimilar. It would also then be possible to improve the security of the master KDC by turning off LDAP. The KDC will be moved to barrett and George will speak to Alastair about what options there are for hardware for the LDAP master. The slave servers which are already upgraded to FC5 will continue to run both KDC and LDAP but these services will be split at the earliest opportunity, when we move to the new building.

  George said that the network had apparently been stable in the last couple of weeks. One switch at Buccleuch Place is showing a warning light and this will be investigated. All the wireless access points at Buccleuch Place have now had their firmware upgraded so that now all wireless access points are running the new firmware. The core switches at Appleton Tower are now running the latest firmware.

• Managed Platform.

  Stephen reported that the original patch to the AFS kernel module had been a little too aggressive and that a more refined patch was now going to be used. The patch to the kernel to support detection of input from USB key boards and mice so that condor jobs could be suspended when a user starts using a machine's console was faulty and a new patch needs to be applied. A new kernel which fixes these problems will be in next Thursday's stable release.

  There are still a few non-MPU rfe services on achilles. These need to be moved elsewhere so that achilles can be upgraded to FC5. Each of these services should be owned by a Unit and their definition needs to be incorporated into the header structure. Either the Infrastructure or Services Unit will take over achilles.

  Support for 64 bit FC5 is now available at the inf level. This is documented on the MPU wiki page on the subject. There is only support for installing from a CD at present (i.e. one can not use a PXE install).

  Stephen has now reached the point in the Inf FC6 Desktop project where one can install an FC6 LCFG managed machine from a CD. He still needs to write notes about this to document a few gotchas.

• Research and Teaching.

  Unfortunately Tim was not present to give his report. Ken invited Rosemary to say something about her recent work within the unit. She said that she had been making further changes to the database tables to support the esteem indicators for the RAE. The new version of ingres (ingres-2.6) had been shutting itself down spontaneously and Rosemary and Tim have been investigating this. They now understand what triggers this behaviour (access to a specific table record as a first query from gurgle) and they have implemented a workaround.

• Services.

  Craig reported that they will be installing a new shezhu server running FC5 at lunchtime today; it will run the shezhu room booking service for Buccleuch Place, Forrest Hill and JCMB.

  The current bugzilla server, kittihawk, is not very stable. It is not fully managed via LCFG and it will not be possible to complete the transition to a fully LCFG managed server before upgrading to FC5 so the transition will continue after the machine has been upgraded.

  It was not possible to schedule the shutting down of the ATABeast in Buccleuch Place (together with the replacement of its disk and turning on of LUN masking) because one of the research groups was having its final review. The work on the ATABeast will probably now be done a week today at 17:30.

  There has been another occurrence of corrupted printcap files on the Appleton Tower print server (the printcaps are regenerated twice a day from data downloaded from LDAP and the corruption presumably occurs during this process). They will be investigating this. Alastair suggested that they should not be rebuilding the printcap entry at the weekend when no support staff are around.

  Craig has started using the SAN; he has transferred some of the data off another server onto the SAN.

  Neil reported a security incident on the group web server caused by a user's wiki being left in a vulnerable state. He will apply a change that he had applied to other servers some months ago.

• User Support.

  Since the last Operational Meeting the User Support Unit had handled 203 new RT tickets (equivalent to about 20 per working day) and resolved 60% of them. There had been a total of 244 tickets (including both new and existing tickets) resolved over the same period.

  It is Morna's last day as a Computing Officer. She will be full time as the Schools Liaison Officer from tomorrow.

  Alison has resolved the problems that she had been experiencing with the migration scripts provided with the latest version of RT. We are planning on introducing the latest version of RT at a quite period after the end of week 12 of the semester (once the teaching has finished). Stephen asked that the information currently in the profile for the RT server should be moved, where appropriate, to a header file.

  The change to the dice-accntmgr rpm which gives a new default value to the afsdir field (referred to in the last minutes) will be in the next stable release.

  Roger has been working on the documentation of the installroot and the LCFG machine installation process in general

AOCB

There was none.