On Tue, 2010-04-20 at 16:41 -0400, James Laska wrote:
Greetings,
Just a heads up, as you know we have a lot more tests running now in our
internal autoqa instance. Anyone watching autoqa-results can see the
activity. Our 3 test clients were having a hard time keeping up,
especially since one system seems to no longer grab a DHCP IP upon
reboot.
As a temporary measure, I've updated the /usr/bin/autoqa script on the
server to not reboot the test systems before or after a test is run on
the client. This allowed 3 systems to work down the backlog of over 200
jobs in one evening.
The patch is attached, I'm not ready to send this out for review yet.
However, I did want to let you know in case you observe any side-effects
from no longer rebooting after every test.
I'd like to get this patch reviewed and into master. I've been running
with this patch for a few months now without error.
Eventually, I can see a need for providing a method to allow tests to
determine whether reboot is needed (before/after). But I don't have any
good ideas on how to tackle that yet, a future exercise. Alternatively,
this could be controlled by way of autoqa.conf. But for now, I'd just
like to disable reboots.
Any objections or thoughts on the previously attached patch (included
below also)?
diff --git a/autoqa b/autoqa
index b4b9097..13e0b1a 100755
--- a/autoqa
+++ b/autoqa
@@ -106,6 +106,10 @@ def schedule_job(controlfile, required_arch=None,
email=None, name=None, dryrun=
cmd = ['/usr/bin/atest', 'job', 'create']
if email:
cmd += ['-e', email]
+ # Don't reboot the client before the test runs
+ cmd += ["--reboot_before=never"]
+ # Don't reboot the client after the test runs
+ cmd += ["--reboot_after=never"]
# some hooks/tests may require special machines
if required_arch:
cmd += ['-m', '*%s' % required_arch]
Thanks,
James