MPU Meeting Tuesday 22nd February 2011

Software Build Farm

Stephen has been adding to the documentation. He's still to write documentation on the build daemons, and an administrator's guide, and a high level programmer's guide.

Porting LCFG to SL6

Work on this has started in earnest.

  • Alastair and Stephen met on the 21st and produced:
  • This time, rather than build up LCFG onto a native install piece by piece, the SL6 port will be bootstrapped using an F13 installroot. This should be quicker.
  • The project will be formally started in the next week with mail to cos.
  • Natural Docs will be used to mark platform support status in headers. The project team needs to meet with Kenny to agree Natural Docs keys for platform support.
  • Alastair has started installing package sets to get base package lists. Stephen suggested using the lists on the disc image in conjunction with yum resolvedep.
  • RC1 is out and has been tagged at 6.0.
  • The final release may appear in the second or third week of March.

Misc Development

build tools and skeleton
Stephen has made more changes to build tools in response to Gordon's bug. Changes were made to the package spec and to lcfg-skeleton. The latter wasn't validating email addresses; now it parses them and generates a sane address. In addition, when the build tools scripts generated CMake they weren't escaping double quote characters in strings: this broke CMake. This has been fixed. Other changes were made to lcfg-skeleton at the same time: for instance it now validates the component name, stopping the user from typing incorrect component names such as "lcfg-foo" or "foo-bar". There's also a new option to pass the --sign option to rpmbuild, fixing LCFG bug 387.
LCFG disaster recovery
this was affected by a bug in nagios monitoring of apacheconf setups using SSL and with multiple interfaces. Its attempts to detect round-robin IP addresses result in a situation where only the primary interface is checked and other interfaces remain unchecked. This bug was not detected until now because sauce is our first such machine not to have its SSL on its primary interface but on a secondary interface. Stephen has come up with a tentative fix but wants to check it further.


  • Stephen has reconfigured the LCFG disaster recovery setup to make the configuration far more robust and less complex.
  • The new F13 kernel is being rolled out.
  • Chris tried using xfree to give canongate a correct nvidia-driver-using X configuration (RT 52016) but got X configuration errors and nouveau drivers that refused to stop appearing. It's possible that nouveau is being started unexpectedly early in the boot process. A way round this may be to add nouveau to the kernel blacklist. Stephen will send Chris details on how to do this. Stephen can also try testing an nvidia driver on his F13 nvidia-equipped test machine.

This Week

  • Alastair
    • Dismantle and reassemble satablade
    • Raise SL6 project to COs
    • SL6 project
      • Create SL6 base lists (inc lcfg and devel)
      • look at upstart under SL6
      • Prepare inf layer headers
      • Produce compilable SL6 profile
      • naturaldocs
    • arrange for latest LCFG install ISOs to be stored on DR server

  • Chris
    • Tell Stephen about two spare 755s (for PkgForge)
    • Look at LCFG slave using ramdisk (using DL180 - fantoosh)
    • Chase RAT about missing f13_64 packages
    • Look at nvidia drivers and RT#52016 (on kelvin or ord)
    • practical submit project

  • Stephen
    • Move PkgForge build off circle onto spare desktops
    • Move circle to AT
    • Change schema version for AFS component (so can upgrade to 8GB)
    • PkgForge docs
    • SL6 mock configuration

-- AlastairScobie, ChrisCooke

Topic revision: r10 - 28 Feb 2011 - 16:48:18 - AlastairScobie
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