Correct errors with amdgpu openmp builders
Details
Diff Detail
Event Timeline
we are trying to get the builder running and seeing the following error.
so we are removing the depends_on_projects clause as an experiment.
Traceback (most recent call last):
- <exception caught here> --- File "/home/stagemaster/buildbot/master/buildbot/process/buildstep.py", line 555, in startStep yield defer.gatherResults(dl)
twisted.internet.defer.FirstError: FirstError[#30, [Failure instance: Traceback (failure with no frames): <class 'twisted.internet.defer.FirstError'>: FirstError[#2, [Failure instance: Traceback: <class 'KeyError'>: 'jobs'
/usr/local/lib/python3.6/dist-packages/twisted/internet/defer.py:151:maybeDeferred
/home/stagemaster/buildbot/master/buildbot/process/properties.py:961:getRenderingFor
/home/stagemaster/buildbot/master/buildbot/process/properties.py:961:<listcomp>
/home/stagemaster/buildbot/master/buildbot/process/properties.py:224:render
- <exception caught here> ---
/usr/local/lib/python3.6/dist-packages/twisted/internet/defer.py:151:maybeDeferred
/home/stagemaster/buildbot/master/buildbot/process/properties.py:463:getRenderingFor
/home/stagemaster/buildbot/master/buildbot/process/properties.py:418:getitem
/home/stagemaster/buildbot/master/buildbot/process/properties.py:145:getitem
]]
]]
From the error message this may fix the problem:
diff --git a/buildbot/osuosl/master/config/workers.py b/buildbot/osuosl/master/config/workers.py index 18d78b28..63bf8a87 100644 --- a/buildbot/osuosl/master/config/workers.py +++ b/buildbot/osuosl/master/config/workers.py @@ -258,5 +258,5 @@ def get_all(): create_worker("arc-worker", properties={'jobs': 16}, max_builds=1), # OpenMP on AMDGPU, Ubuntu 18.04.5, Intel(R) Xeon(R) Gold 5218 @ 2.30GHz with 64GB Memory, 1 Vega20 GPU with 16GB Memory - create_worker("omp-vega20-0", max_builds=1), + create_worker("omp-vega20-0", properties={'jobs': 6}, max_builds=1), ]
Did you consider setting up your own buildbot-master, possibly running on the same machine, to test configurations?
@Meinersbur Thanks. I'm going to try your suggested changes to workers.py. Yeah, I had started to look into trying to setup a buildbot-master server locally, but I ran into connection timeouts trying to access the buildbot site today - may just be my local connection though.
Galina fixed out workers issue, thank you.
David, please abandon this patch.
i will push another to change the settings for : "openmp-offload-amdgpu-project", and : "openmp-offload-amdgpu-runtime",