service record did not updated on time after restart

[Guangya Liu (Jay Lau)]

When using db driver or mc driver for servicegroup, if the service was created before and restart, then the service record will not update on time but need to wait a delay_interval (10s by default) before update.

This will cause VM deploy failed when nova compute restart and user deploy VM immediately with ComputeFilter.

[Russell Bryant (russellb)]

Since you assigned yourself to this bug, I take it that you're working on a fix? I took a quick look at the code and a quick guess is that the "initial_delay" needs to be set to 0 for updating the db.

diff --git a/nova/servicegroup/drivers/db.py b/nova/servicegroup/drivers/db.py
index 5a0869d..a25b525 100644
--- a/nova/servicegroup/drivers/db.py
+++ b/nova/servicegroup/drivers/db.py
@@ -49,7 +49,7 @@ class DbDriver(api.ServiceGroupDriver):
         report_interval = service.report_interval
         if report_interval:
             service.tg.add_timer(report_interval, self._report_state,
- report_interval, service)
+ 0, service)

     def is_up(self, service_ref):
         """Moved from nova.utils
diff --git a/nova/servicegroup/drivers/mc.py b/nova/servicegroup/drivers/mc.py
index a22aa7c..3df5e37 100644
--- a/nova/servicegroup/drivers/mc.py
+++ b/nova/servicegroup/drivers/mc.py
@@ -60,7 +60,7 @@ class MemcachedDriver(api.ServiceGroupDriver):
         report_interval = service.report_interval
         if report_interval:
             service.tg.add_timer(report_interval, self._report_state,
- report_interval, service)
+ 0, service)

     def is_up(self, service_ref):
         """Moved from nova.utils

[Guangya Liu (Jay Lau) (jay-lau-513)]

Thanks Russell, that's exactly my code diff. ;-)

But I did not find a good way to add unit test, and after some checking, seems we do not need to add unit test for this.

Will upload a patch soon.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/47051

[John Garbutt (johngarbutt)]

See new fix:
https://review.openstack.org/#/c/71439/

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/71439
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=afb5cb64b59c982e541cf48de0f126f6e8e2400f
Submitter: Jenkins
Branch: master

commit afb5cb64b59c982e541cf48de0f126f6e8e2400f
Author: Michael Wilson <email address hidden>
Date: Wed Feb 5 18:41:42 2014 -0700

    Change initial delay for servicegroup api reporting

    Initial delay should not necessarily be equal to
    report_interval. Instead make it a reasonable value
    such as 5 seconds.

    This is so the surprising behavior of a service not
    reporting when you start it up doesn't happen if your
    report interval is configured to a larger value.

    For example, in an environment where there are many
    services it may not be optimal to have them all reporting
    every 10 seconds which is the default. You may have them
    report every 5 minutes. It is confusing to start up the
    service and not have it show as available until 5 minutes
    later.

    Fixes bug #1226370

    Change-Id: I85c4146bfa465fffaf89d4c2c74e6cb4ef84cee2