Merge lp:~jk0/nova/lp754983 into lp:~hudson-openstack/nova/trunk

[Josh Kearney]

Try to be nicer to the DB when destroying a libvirt instance.

[Devin Carlen (devcamcar)]

good find, lgtm

review: Approve

[Vish Ishaya (vishvananda)]

lgtm.

review: Approve

[Vish Ishaya (vishvananda)]

is this enough of a performance issue to get into the release?

[Devin Carlen (devcamcar)]

It does look potentially pretty evil. I'd like to see an FFE for this. Very self contained and low risk.

[Rick Harris (rconradharris)]

lgtm. +1 on FFe, very little risk, and much safer.

Long-term:

I'd like to see timeout here so we don't get into a situation where we poll on this for hours, days? on end.

Perhaps we should be using LoopingCall here, in which case, I'd like to see a timeout facility added to it so that we can re-use that throughout the codebase.

review: Approve

=== modified file 'nova/virt/libvirt_conn.py'
--- nova/virt/libvirt_conn.py	2011-04-11 14:17:29 +0000
+++ nova/virt/libvirt_conn.py	2011-04-13 19:12:33 +0000
@@ -372,6 +372,9 @@
                                       instance['id'], state)
                 if state == power_state.SHUTOFF:
                     break
+
+                # Let's not hammer on the DB
+                time.sleep(1)
             except Exception as ex:
                 msg = _("Error encountered when destroying instance '%(id)s': "
                         "%(ex)s") % {"id": instance["id"], "ex": ex}