llvm dev - Apr 2016 - Randon failures in clang-ppc64le-linux-lnt

Hi Bill,

The PPC LNT buildbot is randomly failing on LNT because of timeouts.

http://lab.llvm.org:8011/builders/clang-ppc64le-linux-lnt

I've checked all recent LNT failures and they were spotted on some
processes (lua, sqlite3, tramp3d, expandfft, etc), so I doubt this is
in any was a real time out due to code-gen issues. I imagine they're
all at the edge already and the standard deviation is enough to get it
past the limit often.

Given that this is mostly noise, can you move this bot to the quiet
build master (http://lab.llvm.org:8014/) so that you can continue to
track it's progress without emailing people all the time? Once it's
stable again, you can move it back to the main master.

Thanks!
--renato

I am running more bots on that particular hardware than I would like to 
because of hardware issues on another server I usually use.  I will 
tweak the run settings for it so the timeouts should not occur.  I will 
also keep a closer watch on it.

On 04/24/16 12:56, Renato Golin wrote:
> Hi Bill,
>
> The PPC LNT buildbot is randomly failing on LNT because of timeouts.
>
> http://lab.llvm.org:8011/builders/clang-ppc64le-linux-lnt
>
> I've checked all recent LNT failures and they were spotted on some
> processes (lua, sqlite3, tramp3d, expandfft, etc), so I doubt this is
> in any was a real time out due to code-gen issues. I imagine they're
> all at the edge already and the standard deviation is enough to get it
> past the limit often.
>
> Given that this is mostly noise, can you move this bot to the quiet
> build master (http://lab.llvm.org:8014/) so that you can continue to
> track it's progress without emailing people all the time? Once it's
> stable again, you can move it back to the main master.
>
> Thanks!
> --renato
>

-- 

-Bill Seurer

You can reduce the number of simultaneous jobs run on a given slave (in the
buildbot config) if necessary to reduce load. But if we can't get reliable
results with reasonably short blame lists (reducing the number of jobs
means longer blame lists as builds are run less frequnetly, etc) that's a
problem - such results won't be readily actionable and the results tend to
become noise at that point, lowering the quality bar for all
buildbots/developers.

On Mon, Apr 25, 2016 at 7:48 AM, Bill Seurer <seurer at
linux.vnet.ibm.com>
wrote:
> I am running more bots on that particular hardware than I would like to
> because of hardware issues on another server I usually use.  I will tweak
> the run settings for it so the timeouts should not occur.  I will also keep
> a closer watch on it.
>
>
> On 04/24/16 12:56, Renato Golin wrote:
>
>> Hi Bill,
>>
>> The PPC LNT buildbot is randomly failing on LNT because of timeouts.
>>
>> http://lab.llvm.org:8011/builders/clang-ppc64le-linux-lnt
>>
>> I've checked all recent LNT failures and they were spotted on some
>> processes (lua, sqlite3, tramp3d, expandfft, etc), so I doubt this is
>> in any was a real time out due to code-gen issues. I imagine
they're
>> all at the edge already and the standard deviation is enough to get it
>> past the limit often.
>>
>> Given that this is mostly noise, can you move this bot to the quiet
>> build master (http://lab.llvm.org:8014/) so that you can continue to
>> track it's progress without emailing people all the time? Once
it's
>> stable again, you can move it back to the main master.
>>
>> Thanks!
>> --renato
>>
>>
>
> --
>
> -Bill Seurer
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20160425/325bc951/attachment-0001.html>