Advertisement
Guest User

Untitled

a guest
Oct 24th, 2014
129
0
Never
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
text 1.56 KB | None | 0 0
  1. [10:59:25] <baoli> yjiang5: Joe indicated change of adaptors, node restart by his teammate, etc. I have suspected that after a device is assigned, and if the nova-compute is restarted, the device won't show up in the libvirt's device list. But that doesn't seem to be the case.
  2. [10:59:53] <yjiang5> baoli: I think at least we can change the pci manager, so that if some instance can't get the devices after restart ,like because of hotplug, we should avoid the compute service restart, and at the same time, we can continue talk with Joe of the potential reason.
  3. [11:00:25] alexpilotti (~anorex@79.114.64.150) joined the channel.
  4. [11:00:42] <yjiang5> baoli: yes, I have the same suspect before, but seems that's not ture. Anyway, I will do more experimental locally.
  5. [11:01:23] alexpilotti (~anorex@79.114.64.150) left IRC. (Client Quit)
  6. [11:01:58] <baoli> yjiang5, I agree. You may be able to repro it by doing this: start a VM with 2 PCIs, exit nova-compute, change the whitelist so that it has one device only, restart nova-compute
  7. [11:02:41] alexpilotti (~anorex@79.114.64.150) joined the channel.
  8. [11:03:09] <yjiang5> baoli: I will try to cook a patch to make sure at least the comptue service will be up. Also I will rebase the compute node id patch.
  9. [11:03:34] alexpilotti (~anorex@79.114.64.150) left IRC. (Client Quit)
  10. [11:03:37] <baoli> yjiang5, yes we should fix the fundamental issue.
  11. [11:03:55] rook (~rook@cpe-174-099-018-231.nc.res.rr.com) joined the channel.
  12. [11:04:08] <superdan> baoli: yjiang5: rook is the reporter of the compute-not-starting bug
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement