[webkit-dev] Mac and Chromium Mac EWS

Eric Seidel eric at webkit.org
Wed Jan 25 12:14:16 PST 2012

It's also possible that the NRWT "don't peg the machine" heuristics
are failing for the EWS bots.  You'd have to SSH in to know.  But if
you see it running too many DRT processes (or too many associated
Python processes) we can change the heuristic:

Also, if tests are crashing, we don't account for the CrashReporter in
our heuristic, and it has a bad habit of pegging one or more cores. :)


On Wed, Jan 25, 2012 at 11:56 AM, Adam Barth <abarth at webkit.org> wrote:
> From the logs posted to queues.webkit.org, it looked like the problem
> was that more than 30 tests were failing on those machines.  To run
> the test on these machines, we'll need substantially fewer than 30
> failures.  For example, I think the Chromium Linux EWS has something
> like 5 persistent failures.
> Adam
> On Wed, Jan 25, 2012 at 11:53 AM, Lucas Forschler <lforschler at apple.com> wrote:
>> I did a bit of investigation on the EWS bots.  They seem to be dying rather quickly, and I've been unable to ssh into them.  I connected up to one locally and the cpu was pegged and the machine was unresponsive.  I couldn't even open a terminal session.  I've had to reboot them with a power cycle.
>> We should bring up another queue that runs tests, and one that does not… that way we can debug what is going wrong.  I can put one machine in the trial queue, and leave the other 4 as normal EWS machines, and hopefully we can catch up quickly.
>> Thoughts?
>> Lucas
>> On Jan 25, 2012, at 10:54 AM, Adam Barth wrote:
>>> +lforschler
>>> 2012/1/25 Osztrogonac Csaba <oszi at inf.u-szeged.hu>:
>>>> Adam Barth írta:
>>>>> On Sat, Jan 21, 2012 at 9:32 AM, Osztrogonac Csaba <oszi at inf.u-szeged.hu>
>>>>> wrote:
>>>>>> Something happened with Mac EWS again and now there are 375
>>>>>> stucked patch. Could anyone from Apple check and fix it?
>>>>> Lucas and I have been working to enable running the tests on the
>>>>> mac-ews.  Unfortunately, there are too many failures on the Mac port
>>>>> at the moment (more than 30, at least on those machines).  That
>>>>> prevents the bots from processing patches.  I'm not sure what the
>>>>> current plan is.
>>>> In this case I think we should stop running layout tests on Mac EWS.
>>>> A working builder EWS is always better than a non-working builder
>>>> ans tester EWS. Now the 7-day "Pass" counter of this bot is zero,
>>>> so this bot doesn't work at least a week ago. :(
>>> Lucas, is that what you'd like me to do?
>>> Adam
> _______________________________________________
> webkit-dev mailing list
> webkit-dev at lists.webkit.org
> http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev

More information about the webkit-dev mailing list