nova-compute stops reporting in grenade

[Robert Collins]

Seen in
http://logs.openstack.org/39/70239/1/check/check-grenade-dsvm/6f4b3bf/logs/new/screen-n-sch.txt.gz

jog says
15:38 < jog0> lifeless: it looks like it has happend before
15:39 < jog0> message:"has not been heard from in a while" AND filename:"logs/screen-n-sch.txt"

[Joe Gordon (jogo)]

It looks like n-cpu (http://logs.openstack.org/39/70239/1/check/check-grenade-dsvm/6f4b3bf/logs/new/screen-n-cpu.txt.gz) just stops logging and presumably doing anything at 2014-01-30 23:22:49.183 and then the scheduler starts erroring out at http://logs.openstack.org/39/70239/1/check/check-grenade-dsvm/6f4b3bf/logs/new/screen-n-sch.txt.gz#_2014-01-30_23_26_06_767 because of no heartbeats from the compute node.

Also swift is throwing some cryptic errors around the same time nova-compute stops logging:

http://logs.openstack.org/39/70239/1/check/check-grenade-dsvm/6f4b3bf/logs/syslog.txt.gz

[Joe Gordon (jogo)]

The swift errors are unrelated, and due to the grenade upgrade process.

[Joe Gordon (jogo)]

Its not clear if this is a valid bug, or an issue with https://review.openstack.org/#/c/70239/

[Robert Collins (lifeless)]

This was fixed by 072e4ad20a6a7f4fff3b59eb7ef3f6fef9aa19d1

[Robert Collins (lifeless)]

(Oh, and it was a -very- valid bug :)