![]() |
Apologies for absence
Alison, Alastair, Archie, Roger and Tim had sent their apologies.
minutes of the last meeting.
Toby pointed out that the reference to a slapd component in the Infrastructure report should have been to the openldap component. Rosemary had also pointed out a slight error in the reporting of the work that she had completed
Report from Computing Executive Group
This item was not taken.
Reports from units.
Infrastructure.
George reported that he had run into more problems when upgrading the routers to FC5; there are problems with the way PXE and DHCP leases behave. There was no implicit default dhcp server in the version of dhcp that we are using and this is a change of behaviour from previous releases. Tomorrow, George will attempt to upgrade goodall, the router in Buccleuch Place.
George said that the network had apparently been stable in the last couple of weeks and that there had been no hardware failures.
The upgrade of LDAP had gone through without any significant problems. Decisions still need to be made about how to handle the upgrade of basilisk, the master kerberos KDC and the master LDAP server. One likely possibility is that there will be separate servers for these two services.
Managed Platform.
Stephen reported that we can now support the FC3 platform for a longer period via the inf domain. On FC3 machines in the inf domain there will be no local LDAP, DNS server or NFS imports (only local or AFS file systems) and no firewall holes for these machines. Simon raised some kerberos issues for these machines as if wrongly configured they would be open to a known security hole. There will be no security patches for these machines.
The LCFG master will be moved from achilles to tobermory this Saturday morning. All LCFG data, including the rfe controlled profiles and the subversion repository for headers will be moved. the orders database had already been moved off achilles onto figgy and will eventually be moved to tobermory.
Stephen emphasised that whenever the defaults file for a component changed then it should be rpmsubmitted from both an FC3 host and an FC5 host since external organisations still use FC3 for their LCFG infrastructure.
The test kernel that has been out for a while now seems to have fixed all the problems that George had reported but there were a couple of known issues; one was the reporting of available space on AFS file systems and the other was a file locking bug. Simon said that the first was not a bug but a feature (i.e. it was a deliberate behaviour change) and that the other was a race bug for which a fix was available.
Stephen has made good progress in the Inf FC6 Desktop project. He warned everybody that there are several changes in the behaviour of the version of make that comes with FC6 (thankfully all documented ones) which were not backwards compatible.
Research and Teaching.
Iain reported that there was a new version of Matlab available now, Matlab 7.3 and that this appears to fix many of the X problems affecting the currently used version.
The Informatics Database server has been upgraded to FC5 and OpenIngres and thought that this fixed some of the problems with the previous version of Ingres used on FC3.
Iain had hit an obscure problem that had been preventing hawthorn.inf from communicating correctly with the SRIF SAN out at the Bush estate. They had tried swapping the card, fibre and switch port without success. However it worked finally when the fibre was shortened and a different switch was used. This afternoon IS will be trying to find out why it hadn't worked with the other switch.
The condor pools are both up and running. There is a problem with running a job on the same node as it is submitted from because condor then logs the job submitter off the node. This will not matter once Iain reconfigures the condor nodes so that one can only submit jobs from one specific node (he will just configure that node so that it doesn't execute condor jobs itself).
He is hoping to upgrade the console servers to FC5 and the latest version of the console server software within the next few days. With version 8 of the console server software it will no longer be necessary to have a separate sshconsole script since this version now supports encryption.
Services.
Craig reported that thanks to heroic debugging on the part of Simon and Neil, the problem with LPRng on FC5 has been fixed. The cause of the problem was nss_ldap making unwarranted assumptions about which file handles it owned.
There are still three services that need to be moved to FC5: bugzilla, shezhu and jabber but there are problems porting some of the rpms.
We now have 4x1.6TB available on the SRIF SAN via the IS server. We will be moving the AMI project data onto the SAN from the JCMB SATABeast. We have lost a disk on the ATABeast in Buccleuch Place and will be shutting it down when the replacement disk is installed in order to turn on LUN masking.
User Support.
Ken reported that we were waiting on clarification from the Managed Platform Unit concerning the header files that were allowed in the LCFG profile of an FC3 machine in the inf domain. Stephen said that Alastair would need to comment on this.
Since the last Operational Meeting the User Support Unit had handled 305 new RT tickets (equivalent to about 20 per working day) and resolved 70% of them. There had been a total of 293 tickets (including both new and existing tickets) resolved over the same period.
Morna would be leaving the computing staff at the end of February to become the Schools Liaison Officer. It is planned that the User Survey that she has designed will go ahead, but that the Services Unit would manage it.
Stephen had asked at the last meeting about the state of the computer accounts of former computing staff. Ken reported that he had closed down the accounts of Jeremy, Julieta, Charlie and Sarah. Of most concern had been the removal of the sysman role and Ken confirmed that it had been removed. Simon commented that the admin principals for these former staff should also be removed; Toby will see to this.
Alison has been working on the newest release of RT and had hit several problems in transferring the data from the current system using the scripts provided; there appear to have been some errors in these scripts. She wished to pass on the news that when she gets this version working satisfactorily it should be a lot quicker than the rather sluggish current version.
Roger had made a change to the update_user script in the dice-accntmgr rpm to allow the afsdir value to be set explicitly. He has also supplied a new script create_afsuser which creates an AFS user and enables the associated file space. The updated update_user and the new create_afsuser scripts have been included in the latest version of the dice-accntmgr rpm which will be in the test release next week. Ken said that he had last night built an even newer version of this rpm which had a new version of the Algorithmic data source. The latter now set the default value for afsdir to /afs/inf.ed.ac.uk/user/<first letter of username>/<username> for non-student usernames. Others seemed happy for this to be shipped as well. Simon mentioned that he had a patch that he would send to Ken which added new users to the PTS (protection service).
AOCB
There was none.
Please contact us with any
comments or corrections.
Unless explicitly stated otherwise, all material is copyright The University of Edinburgh |
![]() |