MPU Meeting Tuesday 1st September 2015

Inventory

The orders parser has been tidied up, and it's about to go live.

Alastair has been working on the clientreport component of the inventory. It runs on each machine, gathering data on the local hardware, firmware and software to feed back into the inventory. He's been tackling each aspect (network, monitors, disks, etc.) one at a time. He's blogged about aspects of the work:

Alastair will investigate the idea of making clientreport use separately installable modules, each of which would collect one kind of machine data. This should simplify the gathering of different sorts of data on different machines. A modular architecture should make maintenance easier, particularly the job of changing clientreport to respond to changes in hardware and firmware arrangements over time.

A modular clientreport seems to be the natural tool for discovering and recording firmware versions - the job currently done by firmwarereport on servers. We could also start gathering firmware data from desktops.

LCFG Client refactoring

The main motivation for this project was to switch the client code to a more standard, widely-used, maintained XML parser. Stephen has blogged about the work: He's developing libraries for parsing LCFG XML. There will be three: one to handle LCFG package specifications; a second to handle resources and components; and a third to parse whole LCFG XML profiles, making use of the first two to do it. They're written in C, partly for efficiency and partly to avoid a seemingly inescapable memory leak in XML::LibXML. Using them, an average LCFG XML profile can be parsed in about a fifth of a second, as opposed to perhaps a couple of seconds with the old code.

Misc development

  • Chris is looking for an alternative to nss_db for SL7 Virtual DICE. It was suggested that he talk to Toby about sssd.
  • We talked about pkgsearch and agreed that rather than adapting it for the various repository and package server changes which have happened in recent months, it would instead be simpler for Chris to produce a simplified web front end to yum search and yum info. It will need to cover multiple platforms, but it should be possible to use a separate yum configuration for each desired platform (mock does this).

Operational

  • All of the main KVM servers now have 128GB of memory, and their last SL6.6 upgrade will happen this Thursday.
  • Various servers have been upgraded to SL6.6.
  • OpenAFS 1.6.14 is out, and we should upgrade to it. There's also a new kernel, ditto. We will aim to upgrade the desktops and main user-accessible servers to both of these in time for the start of session. Other machines can be done later. OpenAFS 1.6.14 is in this week's stable release for SL7 DICE machines.
  • Alastair has been looking into how to test the IBM DS3524 storage array with the new FC switch. The idea had been to run two FC fabrics, one with an old switch and the other with a new switch, initially for testing. Since the IBM has two controllers he tried moving all volumes to one controller on the old fabric so that the other controller could be used separately on the new FC fabric. However, it turns out that the IBM just allows you to set a "preferred controller" for volumes. That doesn't give a definite enough separation for this purpose - so it looks as though we won't be able to test the IBM with the new FC switch while the IBM is in service. The new plan is therefore to migrate critical VMs off the IBM to local server disk pools, migrating non-critical VMs on to the IBM if necessary to make room; then wait until both SAN fabrics are running on new switches, take the IBM out of service and move it across to the new fabrics.
  • The IBM is now out of warranty. Alastair has downloaded firmware and software updates, but it's not clear whether or not they'll apply.
  • The improved bridging configuration on the new KVM server azul isn't quite compatible with those on its sibling servers. This means that until we reconfigure those servers and an awful lot of VMs, we won't be able to migrate freely between the servers. We decided to postpone the bridging improvement until SL7, and revert azul back to the same arrangement as the other servers in the meantime (done 2/9/15).

This week

  • Alastair
    • Inventory project
      • continue working through InvProjectWorkFlow
      • finish deploying new order file processing code
      • consider what next can be integrated into existing system, if anything
      • Consider sublocation, macaddr discovery tool (use same code as used for clientreport)
      • Blog articles
      • Continue with clientreport
        • Design module framework - disks, cpus, memory, nics, partitions, disks, RAID - don't have all module code on all machines. Command line choose which modules to use.
        • eventually want to allow for validation at server end, but for now just trust data
        • look at JSON -> postgresql -> only in 9.2 (SL7) and JSONB in 9.4. Not sure how mature it is.
          • for time being, store in JSON in VARCHAR. Script which loads information from reports into other tables will have to decode the JSON. OR we decode the JSON at the CGI end of clientreport, and only store eg one monitor, one macaddr.
    • @home - look at using rsync from site.pkgs instead of mirroring from upstream
    • Remove default pool if ops meeting agrees
    • Add instructions on adding extra disks to an LVM/KVM pool - to KVM host guide
    • Talk to Phil Wadler about machine upgrade - ask for a list of software he requires built
    • Experiment with different window managers under VNC (making the assumption that performance under NX will be similar)
    • Think of a use for 'atom'
    • Understand how NetworkManager works wrt init scripts
    • Flesh out Base SL7 server project

  • Chris
    • Inventory project
    • pkgsearch for SL7
      • reimplement as a yum web front end (yum search for keyword produce an html file of links to cgi to do yum info)
      • Need support multiple platforms
    • Flesh out Base SL7 server project
    • Continue work on VirtualDICE for SL7
    • Groundwork for bugzilla upgrade
    • Liaise with George over iDRAC documentation
    • Check full range of yum repositories is on login servers
    • On azul - revert to br33 (from br0) - record on activities list that we're postponing this scheme until SL7 KVM servers
    • Schedule jubilee 6.6

  • Stephen
    • LCFG client refactor stage 1
      • schedule debrief meeting
    • LCFG client refactor stage 2
      • document C libraries
      • work on C -> perl interface (XSS)
    • Think about PD - Interested in ZeroMQ
    • Finish off window manager selector code
    • Flesh out Base SL7 server project

-- AlastairScobie - 01 Sep 2015

Edit | Attach | Print version | History: r14 < r13 < r12 < r11 < r10 | Backlinks | Raw View | Raw edit | More topic actions...
Topic revision: r12 - 07 Sep 2015 - 16:05:40 - 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