1. guilhem
2. Thorsten
3. Emiliano
4. Brett
5. cloph

 * CI: Thorsten's suggestion in last week's ESC call
   + guilhem: makes sense to me, been using that for Debian builds and tests
     - downside is that we'll loose homogeneity (throw away disk overlays won't 
work on metal hardware)
     - counter proposal: use namespaces or containers
       . Thorsten: easier to use jenkins libvirt plugin
       . overhead of machine-level isolation might be bearable, there is
         no ccache so powerful fast enough VMs
       . Thorsten: not ideal on windows, but can do that for linux
         . Brett: linux subsystem can be an option on windows
     - cloph: like the idea of using VM and throwaway disk overlays for windows
       . windows is the only platform with leftover problems
     - 5 active metal windows boxes (incl. 3 at manitu, 48 cores 62G RAM)
       . cloph: perf is misleading, tb88's underlying metal is newer and
         doesn't have near these resources, yet in performs well
     - AI cloph to deploy a test setup, but keep tb7[237] for now;
       deadline when windows server 2012 is EOL some time next year
 * (Proposal by Florian, given the timeline) Decommissioning of two
   hypervisors and ~7 machines from the rack by the end of the year
   + juttan, kuleha, lilith, maggie (tb7[2356]) — all 4 1U
     - migrate + decommission linux boxes (lilith, maggie), keep the other 2 
for now
   + navi, okul, pollux (tb7[789]) — all 3 1U
     - migrate + decommission tb78 and pollux, keep tb77 for now
   + berta — 4U
     - keep it for now
   + charly, dauntless, excelsior, falco — 4U(?)
     - process these in order
   + pitr — 2U
     - migrate + decommission
   + miranda — tower (4U?)
     - migrate + decommission
   → 2+2+1+1 = 6 + 1 (2 if possible) hypervisors, nearly half the rack (~18U)
   + AI: guilhem to deal with the hypervisors, pitr, and miranda; cloph to take 
care of the tbs
 * Next call: Tue Nov 15 at 17:30 UTC (18:30 CET, note the DST)


To unsubscribe e-mail to:
Posting guidelines + more:
List archive:
Privacy Policy:

Reply via email to