Apologies for absence
Carol and Craig had sent their apologies.
minutes of the last meeting.
These were accepted.
Report from Computing Executive Group
This item was not taken.
Reports from units.
Appleton Tower level 3 had been handed back to the University last week and level 4 should be handed back tomorrow. The technicians had done sterling work in levels 6, 7 and 8 to get all the networking patched to the switches and the switches configured. They will shortly be doing extra work in installing panels and base units into the flip desks that are to be installed on levels 3 and 4.
The next work within the Appleton Tower after the above work is completed will be in the basement. Currently this work is at the design phase. In connection with the preparatory work for the basement it is likely that there will be interruption to power in the basement (affecting our own and IS's server rooms) during some weekend near to the start of semester. George will be trying to find out the exact date as soon as possible.
The networked power block (an APC Switched Rack Power Distribution Unit) that had been installed in the new rack in the Appleton Tower server room is now almost fully utilised and is drawing about 3.5A. An additional PDU is in the queue to be ordered.
The unit will shortly be introducing two new infrastructure servers for Appleton Tower. They are kubelik and jarvi - both Dell PE 1950s running FC6. They will be made the primary external and secondary routers for Appleton Tower tomorrow and will take on their other functions shortly thereafter.
After receiving several complaints about poor wireless coverage George has investigated the situation and discovered that several of the wireless access points were suffering from a firmware bug which allowed the access point to lose part of its configuration. George has now started a programme of upgrading the firmware on all the wireless access points.
Following on from the problems with the JCMB server room air conditioning reported at the last meeting, George and the technicians have taken steps to improve the effectiveness of the air-conditioning in the server room at the same time as improving the working environment in the adjacent technicians work room by swapping floor tiles with inbuilt grills from the work room with plain floor tiles from the server room. This has raised the temperature within the work room to a more agreeable level and at the same time reduces the occurrence of hot spots within the server room. In doing this work some floor grill tiles were also borrowed from the room that has recently been handed over to the College for its use. Dave has also contacted the Works Department about checking the ceiling tiles and location of baffles, and also chased up the fixing of the second of the backup air conditioning units in the ceiling.
Neil reported that the air conditioning monitor in the Appleton Tower server room appears to be faulty since the alarm has been triggered on a couple of occasions recently even though on inspection the room seems to be adequately cooled. There has been one occasion when both the temperature low and temperature high alarm states have occurred simultaneously. In addition, Ian had also commented that he had found the server room overheating on one occasion quite recently. Neil will report the situation to Dave Hamilton.
Stephen reported that they were in the middle of introducing two new LCFG slave servers mousa and trondra. Currently they had the aliases lcfg2 and lcfg4 respectively but these aliases would shortly be swapped with those currently applied to kipper (lcfg1) and boreas (lcfg3). The aliases lcfg and lcfghost already point at mousa. The kipper and boreas servers will be decommissioned sometime next week.
Alastair had finished the QLogic Fibre Channel support for FC6 which allows servers with fibre channel cards to be ported to FC6. The QLogic software for configuring fibre channel switches has not been ported to FC6 and we may well choose to purchase the more user friendly replacement software for this task which has a GUI. The unit believe that all server support for FC6 has now been done.
There have been a couple of problems with graphics cards under FC6:
The most serious is the problem affecting the on-board graphics card in the new Dell GX745 under FC6. It is impossible to switch from the X display to a pseudo terminal without fatally corrupting the contents of the video memory, necessitating a reboot of the machine. This appears to be mainly fixed in FC7 but the task of back-porting software from FC7 to FC6 in this case is too great and there is therefore nothing we can effectively do about this.
The second problem affects 3 to 4 year old machines with an additional NVidia graphics card. The device driver for these cards has not been ported to the latest release of the kernel (the legacy driver had worked with an earlier version of the kernel). Any machine that has an additional NVidia graphics card and which needs to be upgraded to FC6 will require a newer graphics card before this can be done.
Research and Teaching.
All beowulf cluster nodes are now using an external LDAP server (seville) instead of a local LDAP server, and as a consequence are much more stable even under heavy load. In making the change from a local LDAP server to an external one a couple of unexpected behaviours were noticed: amd has its LDAP server location hardwired as localhost (this will need to be changed so that it is configurable), and there is a bug somewhere in the bash defenv setup that breaks if the LDAP server resource is a list of servers instead of a single server name.
Firefox 2 is currently on the develop release of FC6. Any running Firefox 1.5 will have to be killed from the command line if it is running when Firefox 2 is installed. We will need to warn users about this issue.
Sicstus3, in parallel with Sicstus4 and antlr are now available in the FC6 stable release.
Iain has just got Condor working on FC6 and is now doing further tests. Once this is completed the condor pools will be switched from being KB and central to being FC5 and FC6 with flocking turned off.
Tim asked George about the availability of Cosign under FC6. George said that he would speak to Toby about this.
Neil reported that a mono laser printer from the level 5 support office in Appleton Tower had been installed on level 8 as a temporary solution to the new occupants' printing needs.
The link between our machines and hawthorn seems to have been re-established. George and Sam Wilson will need to meet to sort out a mutually agreeable solution to the routing problem since there are conflicting requirements from Informatics and IS. Sam would like all SRIF network traffic to be routed via one of the IS routers so that the firewall that they have set up can't be by-passed. However if we sent all SRIF network bound traffic from our hosts over our internal networks it would seriously impact the performance of some of our routers.
A disk in the Buccleuch Place ATABeast has failed. The ATABeast is no longer under warranty with Nexsan and so the failure has been reported to NCE with whom we have a support contract. They will be shipping a replacement disk to us shortly.
The unit recently came across a bug in the rsync component. If the rsync component is stopped (or restarted) it kills all running rsync processes (unnecessarily). This component now belongs to the Services Unit so they will look into fixing this behaviour.
The unit have completed a move of data from scargill to mcgahey for the data mining group prompted by a failure of a disk on scargill. This job would have been done by Roger but he is now on annual leave.
Two new servers, cigar (a Dell PE 2950) and mullo (a Dell PE 1950), have been installed. Mullo has a QLogic fibre channel card installed and will be used as a test linux home directory server.
The unit have been looking at plone for the web cms (content management system). Plone 2.5 is available for FC5 and FC6 via the extras repository. Plone 3.0 was released yesterday and has features already built into it as standard which have to be added to plone 2.5. However we are dependent on some recent work for plone that was targeted at the plone 2.5 release so we may still start with that release. It should be fairly straightforward to migrate data from a plone 2.5 based service to a plone 3.0 based service.
Neil has modified the buildquotas script in the dice-quotas component so that it is now possible to have role based quotas that are additional rather than absolute.
The porting of local LaTeX style files to FC6 is now essentially complete but there are outstanding problems with style files using Perpetua fonts. Neil commented that he thought it should be possible to make the use of the Perpetua fonts conditional and to normally use an alternative open source font in their stead.
The plan for the installation of machines on levels 3 and 4 of Appleton Tower is:
If all goes well levels 3 and 4 will be fully usable by Monday 3rd September.
Ken reported that he had finished most of the clean up the quotas file and had rationalised the way the quotas were recorded in the file so that it was easier to see what the situation was.
Since the last Operational Meeting the User Support Unit had handled 147 new RT tickets (equivalent to about 15 per working day) and resolved 55% of them. There had been a total of 154 tickets (including both new and existing tickets) resolved over the same period.
Ken reported some figures showing resolution rates against age of recent RT tickets:
|Ticket Age D||Percentage Resolved (or rejected)|
|1 week < D < 2 weeks||70%|
|2 weeks < D < 1 month||79%|
|1 month < D < 3 months||91%|
The total number of accounts (not including temporary ones) which have an AFS home directory is now 190 (an increase of 9 in the last fortnight).
About 112 machines had been upgraded/installed with FC6 in the last two weeks. Ken presented the following figures for numbers of FC6 and FC5 machines:
There was none.
Please contact us with any comments or corrections.
Unless explicitly stated otherwise, all material is copyright The University of Edinburgh