Operational Meeting

Apologies for absence

Craig, Rosemary and Tim had sent their apologies.

minutes of the last meeting.

These were accepted.

Report from Computing Executive Group

In answering questions Alastair reported:

There were no further details about when Mike would talk to computing staff about accommodation in the Informatics Forum.

Alastair will be starting to place information about the Appleton Tower building timetable on the DICE wiki at https://wiki.inf.ed.ac.uk/DICE/AppletonTower

After some mix up it had now been agreed that we would vacate the Systems Design lab (JCMB 1206c) by the end of May, at the latest. The data wiring for that room and the adjacent north machine rooms lab (JCMB 1206b) are situated in that room. We will be vacating the other labs in that area by the end of June.

Reports from units.

• Infrastructure.

  The planned master KDC server barrett is now running as an FC5 KDC slave server. The Unit will be discussing details of the switch over to this as the KDC master in their meeting tomorrow. The new LDAP master server hardware has been ordered.

  The JCMB wireless access points have been working satisfactorily using power over ethernet for the last few weeks. They draw very little power and the switch could, if necessary, provide power to many times more access points than it is currently.

  The network has essentially been stable over the last two weeks. The only problem was caused by some issues associated with a reboot of solti yesterday morning: bogus entries in one of the spanning maps had hung the machine as it was booting. George has implemented a workaround (the spanning maps are no longer being processed at boot time).

  Iain reported that he would be upgrading the console server at Appleton Tower this afternoon and would probably do the two at Buccleuch Place tomorrow. That would just leave the one at Forrest Hill, which might be awkward because of its remote location relative to the racks and the consequently trailing leads between the console server and the servers.

  George announced that it was two years since we had started capturing images of the Crichton Street car park/Informatics Forum building site and that he had created a script to select images four times a day from periods in the past spaced at about 45 days apart to illustrate the history of the building of the Forum.

• Managed Platform.

  Stephen reported that the new kernel and new AFS kernel module that was currently in the testing release should fix all the known problems with AFS locking, detection of USB connected mice and keyboards (relevant for ensuring condor suspends jobs when someone is using the console) and the stability issues for the routers. If all goes well it will be in the stable release tomorrow. If that goes ahead then it will be necessary to reboot all FC5 desktops as soon as possible thereafter. If it is not shipped as a stable release then it will be important to back port the USB fix to the current stable release kernel.

  FC6 software releases will start next week. Installation of a stable FC6 release from CD will be fully supported and PXE installations of FC6 will be possible from the week following. FC6 will be tested on all relevant desktop hardware (approximately Dell GX240 upwards) and a support matrix of what is supported on what hardware will be maintained and published. It is likely that within the next few weeks it should be possible to have a usable FC6 desktop (using dice level headers as opposed to the inf level headers).

  There are no longer any RedHat 9 machine profiles (all former rh9 profiles have been converted to using the self-managed, junk or unallocated headers). It is now only the solaris headers that are still being maintained via rfe.

  All of the servers that are the responsibility of the unit are now running FC5.

• Research and Teaching.

  Archie has been having an interesting time trying to use the Microsoft Vista software licences sourced from the Microsoft Select Scheme. About 15% of the licences won't authenticate to Microsoft's authentication server at installation. It has also been impossible to use them in the project lab because of restrictions on the minimum number of hosts that must be licenced and the Windows server software required.

  Graham has just about finished his work on the HiGraph project and has taken over responsibility for the postgresql component

  Once a new kernel with the USB fix is out, the live condor headers can be modified to prepare machines for becoming condor clients. Initially all lab machines will be used (but excluding those in quiet labs - specified in a resource that holds a list of rooms). At present it is proposed that staff should have the option of including their desktop in the condor pool (opt in), but this may be replaced by a more sophisticated arrangement whereby staff can move their machine in or out of the pool themselves.

  John reported that he had finished building the backup matlab licence server (to be used as either the research licence server or the teaching licence server). Since the machine used as this backup licence server will be underused otherwise it will also be introduced as the pgresearch postgresql database server. John also said that he had finished his work connected with the COBrA-CT project, and the video-conferencing facilities for I-Rescue and the commercialisation centre planned for Appleton Tower levels 7 and 8.

• Services.

  Neil reported that a problem with the Appleton Tower print server had arisen again, with a different error message but otherwise similar symptoms to before (the CPU was 100% busy). There has not been time to diagnose the problem and it is not repeatable. Restarting LPRng fixes the problem so a work-around has been implemented whereby LPRng is restarted every morning via a cron job.

  Achilles (formerly managed by the MPU) has now been upgraded to FC5. The bugzilla service has been moved temporarily from kittyhawk to smiths. It will eventually be moved to maelcum, once any issues with running two apache and two MySQL servers on the same machine are resolved (maelcum is already running the shezhu resource booking service).

  The jabber server boogaloo still needs to be upgraded to FC5 but that is something that Simon said that he would do.

  There has been a second disk failure in the SATABoy in Appleton Tower in as many weeks; however the disk has now been replaced and we have a full complement of spares.

  Neil has moved all his mail folders from mail.inf.ed.ac.uk to staffmail.ed.ac.uk and has documented his experience as well as a lot of other related pieces of useful information on the DICE wiki under Staffmail Notes . He used imapsync to manage the transfer and synchronisation of folders on the two imap servers; this should be available on all machines in a future stable release (possibly tomorrow). Cross-domain authentication is not yet fully supported but Neil has got his machine doves configured to do this.

  There has been some work done on mirroring data via har over NFS to the SRIF SAN but it is not yet complete. Alastair asked about rsync'ing data from hawthorn (which is based at the Bush estate and is directly connected via fibre channel to the SAN).

  Neil mentioned a talk given at a recent IS Web Clinic which he and Gordon had attended. There was a presentation on MIS project OTH049 University Directory managed by Rose Dewar. It is a sub-project of the University Web project. Looking at the Project Issues & Control Log it would seem that a lot of the data will be sourced from IDMS but that it will also be possible to add functional contact information such as the inf.ed.ac.uk mail domain's hod alias (which is not available from IDMS). During Neil's description of the talk it was not clear what the relationship was between this project and IDMS and for this reason it was agreed that it would probably be discussed at CEG

  At the same meeting there was a talk by someone from Finance extolling the benefits of using an IS event booking facility. We should perhaps consider this for the few occasions when we require such a facility within the School.

• User Support.

  Ken omitted to give the following figures in his report but provides them now:

  Since the last Operational Meeting the User Support Unit had handled 184 new RT tickets (equivalent to about 19 per working day) and resolved 61% of them. There had been a total of 192 tickets (including both new and existing tickets) resolved over the same period.

  Ken had written a script to warn former users that their account was going to be deleted and he has passed on details of the message that is sent to Alison. She will consult with Lindsey about whether the text needs to be modified.

  Ken has also written a script to generate deletion scripts to be run as root on the file servers. The script asks for confirmation only in unusual/unexpected cases.

  Alison has spoken to the current LaTeX support person and he has now submitted his PhD and is waiting for his viva to take place. Until this happens he is happy to continue supporting LaTeX. Alison will mail out the job advert nearer the time of the viva.

  Ken has continued to investigate the replacement orders/inventory system. He will be presenting his plan at the next Development Meeting and may, depending on the availability of another speaker, being giving a talk on the subject in the second half of the meeting. He has started to write a script to extract inventory information from (a copy of) the School database and insert new records and update existing records with data from the XML format order files.

  Alison still plans to upgrade the RT server to the latest version of RT on Wednesday 4th April.

  Ken had mailed all computing staff last Friday to let them know about a file that is included in several web pages that informs users of the state of systems. The file is in the repository of the Informatics web server in the location systems/support/alertmessage.inc (visible as raw HTML at http://www.inf.ed.ac.uk/systems/support/alertmessage.inc) which is included as part of the following three pages:

  • http://www.inf.ed.ac.uk/systems/
  • http://www.inf.ed.ac.uk/systems/support/
  • http://www.inf.ed.ac.uk/systems/support/form/

  Stephen mentioned that when accounts are being created with AFS home directories the Unix permissions on the AFS home directories were not being set with the consequence that they appear to be world writable (even though they aren't because of the AFS access control lists). Ken will look into this.

AOCB

There was none.