Unity autopilot test container timing out/dying on ati

[Łukasz Zemczak]

Ok, it's hard to say what's going on, but... we're having problems with otto autopilot testing of the unity stack, only on the ati machine.

13:55 < sil2100> didrocks: sooo, mhr3 did some research and we think we're encountering
                 the same recordmydesktop issue as before (which caused jibel to disable
                 it) - but this time only on ati
13:56 < sil2100> didrocks: mhr3 ran unity AP tests on his ati machine in his office and
                 there was all fine, ~17 failures he had - he's re running it
13:56 < sil2100> didrocks: we had 13 failures on intel

The system stops with this message, hanging:
E: Test failed to run in 7200 seconds. Aborting!

There seem to be issues with attaching to the container. Best to see the latest run:
https://jenkins.qa.ubuntu.com/job/autopilot-saucy-daily_release/575/label=autopilot-ati/console

Of course, this might be also an unity issue... who knows?
This bug is HIGH PRIORITY! It's blocking the unity releases.

[Stephen M. Webb (bregma)]

/var/log/syslog: Jul 23 10:32:49 saucy-i386-20130723-0916 kernel: [15847.865337] unity-panel-ser invoked oom-killer: gfp_mask=0x50, order=0, oom_score_adj=0

That certainly doesn't sound right.

[Jean-Baptiste Lallement (jibel)]

I increased the memory limit of the testing environment from 2.5G to 5G. We'll see if it reproduces with this setting but it is a rather high setting and not all the test machines have more than 4G of physical memory.

[Łukasz Zemczak (sil2100)]

Seems to be an otto issue.

[Jean-Baptiste Lallement (jibel)]

That's really an issue in:
1. recordmydesktop that constantly allocate/deallocate large pages of memory
2. autopilot that starts recordmydesktop for all the test cases, which for unity means starting and stopping rmd 500 times or so.

Closing otto task as there is nothing we can do here. The limitation is on purpose to prevent against a test to crash the host system if it uses too much memory.