MPU Meeting Tuesday 7th May 2013

Inventory

Alastair has been doing more prototyping and discovering more problems with the existing inventory infrastructure.

Location finding now works better. Alastair and George have agreed a data format for George's automated MAC history information gathered from the network data - MAC address, port, time, indexed by MAC address. Any multiple reports for the same machine are combined and only the most recently timed one is used.

Machine descriptions will now be more uniform - previously descriptions of the same model could differ depending on whether the machine was DICE or self managed.

Several other schools are using OCS Inventory, which automatically reports back inventory information from clients. It covers a wide range of platforms and it's well thought of. It seems impractical for us to use it on all of our machines but we may well use it to populate the DICE client reports.

The inventory will stay at arm's length from Theon, in its own database, with data being fed to Theon by means of conduits. This should make for better security and safer development.

Login Logs

We now have a draft Data Protection statement which seems complete enough to to allow Stephen to go ahead and roll out Log Cabin.

Energy Savings on DICE Desktops

David Sterratt has noticed, in the BuzzSaw sleep report, machines apparently sleeping or waking twice or more in a row. This seems to happen fairly frequently. Chris has taken a look and found a case in which several wake events were logged correctly in the machine's syslog but weren't received by the central log host. The infrastructure unit now has the details.

David Sterratt has produced his first graph of DICE desktops' collective sleep behaviour. It's pretty encouraging: graph of DICE sleep rates 2013−04−25 to 2013−05−02 This probably underestimates the amount of sleep, but it includes only those machines which slept at least once during the period. We have about 800 active DICE machines so the majority of them are not yet sleeping.

This could be helped by switching to the sleep component version currently in beta testing. It seems to be performing well, so Chris will deploy it, though not until after the 21 May exam.

DICE virtual image with teaching software for students - trial

This project hasn't seen much activity since the last meeting. There are two aspects to look at next:
  1. Making LDAP-derived information available in an insecure VM that's probably going to be running outside our firewall. The Infrastructure Unit is taking a look at authenticated LDAP access from outside the firewall, and at LDAP on clients in general, but in the meantime Chris will try Stephen's idea for dumping necessary LDAP info (regularly on a server into nssdb in ACL-protected AFS space, from where it can be pulled down onto a VM after the user has acquired suitable DICE Kerberos and AFS credentials).
  2. Make the test VM importable by others and make it available for downloading by interested parties.

LCFG Client refactor - code cleanup

Stephen has been blogging about his progress.

The code has now all been tidied enough to make perlcritic reasonably happy.

The next stage is to make the behaviour more properly modular where possible. Logging has been pulled out into a logging module. There will also be a utilities module. The tangle of calls between the existing modules should then be simpler and clearer. Global variables are now in an object, with methods, which seems a lot clearer than before. The contexts code was formerly in two places with lots of duplication, but is now in a contexts module, to which Stephen has added locking. As a result the setctx script is now far smaller and consists largely of calls to the contexts module.

The next stage will be to get rid of hard-wired paths in code, possibly using a module which will work out and supply the correct location for a file. The current code has multiple separate places where suitable file locations are worked out.

Other aspects still being pondered are regular DNS lookups for LCFG servers and the reliable opening of a network port to listen for notifications.

Operational

  • The hardware for the new AT-based KVM server has arrived. Chris will deploy it.
  • Carol has finished her audit of VMs on the KVM servers. She found several VMs that could be deleted, two of which have already gone. She's also added some useful state information for each VM to the KVM servers' wiki pages, which are linked from SimpleKVMDocs.

This Week

  • Alastair
    • Educate individuals about inappropriate KVM guest sizes
    • Start Inventory project diary
    • Inventory project
    • RT tickets
    • Review outstanding LCFG bugs
    • Discuss mail logperms with Neil (make default secure at LCFG)
    • Check northern/piccadilly are correctly bonded (re INF unit suggestion)
    • Check new kernel performance on juice (more methodically this time)
    • T1 figures due

  • Chris
    • Ask Perdita how she fixed her Xemacs problem - and write a blog article, if appropriate
    • RT tickets
    • Virtualised DICE image - looking at local identity caching
    • Look at RT:61762 - perhaps a later 279 kernel available
    • Review outstanding LCFG bugs
    • Deploy new sleep everywhere (asking Alison first if all on-line exams are finished)
    • Deploy new VM server (in AT)
    • T1 figures due

  • Stephen
    • RT tickets
    • Deploy and publish user log viewer (and blog article)
    • LCFG code cleanup
    • Talk to Graham about limits component
    • T1 figures due

  • Carol

-- AlastairScobie - 07 May 2013 -- ChrisCooke - 08 May 2013

Topic attachments
I Attachment Action Size Date Who Comment
pngpng sleep-stats.png manage 24.9 K 08 May 2013 - 08:16 ChrisCooke graph of DICE sleep rates 2013−04−25 to 2013−05−02
Topic revision: r5 - 14 May 2013 - 08:42:43 - ChrisCooke
 
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