Skip to content

[btlr] testing: use NUM_TEST_WORKERS env var #4030

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 6 commits into from
Jun 12, 2020

Conversation

tmatsuo
Copy link
Contributor

@tmatsuo tmatsuo commented Jun 9, 2020

Experimenting with different number of workers.

lint, py-2.7, py-3.6:
default (I think it's 6 on Kokoro)

py-3.7:
12

py-3.8:
20

@tmatsuo tmatsuo requested a review from kurtisvg June 9, 2020 21:50
@tmatsuo tmatsuo requested a review from a team as a code owner June 9, 2020 21:50
@googlebot googlebot added the cla: yes This human has signed the Contributor License Agreement. label Jun 9, 2020
@tmatsuo
Copy link
Contributor Author

tmatsuo commented Jun 9, 2020

Seeing lots of failures in py-2.7.

Command python /workspace/appengine/standard/download-appengine-sdk.py /tmp/tmp1midmko3 failed with exit code 1

@tmatsuo
Copy link
Contributor Author

tmatsuo commented Jun 9, 2020

Build time

lint: 7 min, 54 sec
py-2.7: 19 min, 34 sec
py-3.6: 1 hr, 9 min, 44 sec
py-3.7: 1 hr, 43 min, 28 sec
py-3.8: 1 hr, 11 min, 38 sec

I think we can just go with the default.

I'll fix some flakes then try the full build again.

@tmatsuo
Copy link
Contributor Author

tmatsuo commented Jun 10, 2020

Now it temporarily includes #4048

All the jobs are just using the default value.

@tmatsuo
Copy link
Contributor Author

tmatsuo commented Jun 10, 2020

Build time without setting NUM_TEST_WORKERS (defaut values):

lint: 10 min, 36 sec
py-2.7: 21 min, 57 sec
py-3.6: 1 hr, 35 min, 33 sec
py-3.7: 2 hr, 7 min, 50 sec
py-3.8: 1 hr, 17 min, 8 sec

The failure in py-3.8 build is reported at #4054

The builds were rather slow, it might be because of some retries, but it might be because of low number of workers.

The last periodic builds are (btlr is using fixed 30 workers):

lint: 9 min, 38 sec
py-2.7: 22 min, 12 sec
py-3.6: 59 min, 5 sec
py-3.7: 1 hr, 1 min, 56 sec
py-3.8: 1 hr, 0 min, 39 sec

I'll try 10 workers next.

@tmatsuo
Copy link
Contributor Author

tmatsuo commented Jun 11, 2020

Build time with 10 workers:

lint: 9 min, 24 sec
py-2.7: 20 min, 42 sec
py-3.6: 1 hr, 28 min, 24 sec
py-3.7: 1 hr, 26 min, 5 sec
py-3.8: 1 hr, 18 min, 41 sec

There were no weird error during pip install.
Maybe we can settle to 10 for now.

@tmatsuo tmatsuo removed the request for review from averikitsch June 11, 2020 20:26
@tmatsuo
Copy link
Contributor Author

tmatsuo commented Jun 11, 2020

Seeing a new failure in py-2.7 build:

E     File "/workspace/appengine/standard/migration/incoming/.nox/py-2-7/lib/python2.7/site-packages/rsa/key.py", line 1
E   SyntaxError: Non-ASCII character '\xc3' in file /workspace/appengine/standard/migration/incoming/.nox/py-2-7/lib/python2.7/site-packages/rsa/key.py on line 1, but no encoding declared; see http://python.org/dev/peps/pep-0263/ for details

@tmatsuo
Copy link
Contributor Author

tmatsuo commented Jun 11, 2020

I'll update the number of workers for the py-2.7 build to 1. The build is already fast enough without running in parallel.

@tmatsuo
Copy link
Contributor Author

tmatsuo commented Jun 11, 2020

It seems like the failures in py-2.7 build are deterministic.

@tmatsuo
Copy link
Contributor Author

tmatsuo commented Jun 11, 2020

https://pypi.org/project/rsa/4.2/#description

Version 4.0 was the last version to support Python 2 and 3.4. Version 4.1 is compatible with Python 3.5+ only.

I think we need to pin rsa-4.0 for some projects.

Update: The new google-auth handles this.

@tmatsuo
Copy link
Contributor Author

tmatsuo commented Jun 12, 2020

Some tests still fail in py-2.7 build.

@tmatsuo
Copy link
Contributor Author

tmatsuo commented Jun 12, 2020

Build time with 10 workers:

lint: 10 min, 0 sec
py-2.7: 21 min, 37 sec
py-3.6: 1 hr, 18 min, 40 sec
py-3.7: 1 hr, 23 min, 32 sec
py-3.8: 1 hr, 20 min, 37 sec

There was only one failure in healthcare, which should be fixed by #4064

@tmatsuo
Copy link
Contributor Author

tmatsuo commented Jun 12, 2020

Restarted py-3.7 build

@tmatsuo tmatsuo merged commit 2efc5fc into GoogleCloudPlatform:master Jun 12, 2020
@tmatsuo tmatsuo deleted the btlr-workers branch June 12, 2020 20:08
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
cla: yes This human has signed the Contributor License Agreement.
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants