On Wed, Sep 14, 2016, at 10:01 AM, Sean Dague wrote: > On 09/14/2016 12:06 PM, Roman Podoliaka wrote: > > Hmm, looks like we now run more testr workers in parallel (8 instead of 4): > > > > http://logs.openstack.org/76/335676/7/check/gate-nova-python34-db/6841fce/console.html.gz > > http://logs.openstack.org/62/369862/3/check/gate-nova-python27-db-ubuntu-xenial/2784de9/console.html > > > > On my machine running Nova migration tests against MySQL is much > > slower with 8 workers than with 4 due to disk IO (it's HDD). When they > > time out (after 320s) I see the very same TimeoutException and > > IndexError (probably something messes up with TimeoutException up the > > stack). > > Yes, by default testr runs with the number of workers matching the # of > cpus on the target. I think all our cloud providers are now 8 cpu > guests. So unit / functional tests are running 8 way. That's been true > for quite a while IIRC. OSIC is 4vcpu and the others have 8vcpu. Clark