AFS weekly meeting

Monday 14 Aug 2017, 14:00 → 15:00 Europe/Zurich

31/1-012 (CERN)

AFS service/operation meeting

present: Dan, Cristi, Kuba, Massimo, Jan

Issues:

• still draining/rebooting
• afs940 rebooted due to Hypervisor hardware issue (memory) - no NO_CONTACT raised (too short or "pooled"?)
  • intervention scheduled, afs940 is now also drained)
• afs268 got BOINCed, auto-restarted (good) but this did not clean up the pending "vos" commands (on afscron1, afsadm2, machine itself) - bad. Done manually, may investigate if this happens again, procedure usually signals the clients to go away, or to resume activity...
• AFS FIM change (C5, gone live today) - users can as to delete AFS account and keep the UNIX entry in LDAP. Well-hidden, supposedly has "are you sure?"-question. Makes more sense once the interface is redesigned, currently still in "Linux/LXPLUS".
• ABS: Kuba implementd a split between "temp file area" and "restore partition" to split the IO ops (in retropective, could have had the same via a symlink for {{/vicepa/abs_restore}}. Even with this, afs900 is still dog slow for "restore" - should go to high I/O ops CEPH, or back to physical machine.
  • Dan recommends to have the "udev" rule on all virtual AFS servers - https://its.cern.ch/jira/browse/AFS-458