> > Again I found two hanging tests:
> >
> >
http://autoqa-stg.fedoraproject.org/results/4507-autotest/virt27.qa/debug/
> >
http://autoqa-stg.fedoraproject.org/results/4508-autotest/virt24.qa/debug/
> >
> > I aborted them and also all the tests that were queued since
> > then.
>
> There is this setting in global_config.ini:
>
> job_timeout_default: 72
> job_max_runtime_hrs_default: 72
>
> The value is in hours. I tried to change it locally to one hour and
> it aborts a hanged job correctly. I intend to set these values on
> our staging (and production, if no one objects) server:
>
> job_timeout_default: 1
> job_max_runtime_hrs_default: 2
>
> I suppose the first value sets the timeout from the last message
> received from the client, the second value sets the timeout for the
> whole runtime (I might be wrong, I found no documentation apart
> from
> [1]).
>
> I'll also mention that in our "Install autotest" wiki page.
>
> That should help a bit.
>
>
> [1]
http://autotest.kernel.org/wiki/GlobalConfig#AUTOTEST_WEB
Hmm, after I set this up on the staging server, some tests were
aborted while queued. I'll talk to Lucas about the exact meaning of
those options.
. The options will also be
described inside the config file in then next autotest release. I decided to leave
job_timeout_default to the default 72 hours, and set job_max_runtime_hrs_default to 2
hours. I'll apply this change to staging and production server. That should protect us
from hanging jobs.
I also documented the recommended value here: