Skip to content

Conversation

@djkirkham
Copy link
Contributor

Fix to get the tests passing while we investigate why there are now so many time-outs.

@djkirkham
Copy link
Contributor Author

Closes #2752

@djkirkham
Copy link
Contributor Author

The tests do still time out sometimes, but if the --process-timeout value is set too high Travis itself times out. I think this is the best we're gonna get.

@lbdreyer @pp-mo

@lbdreyer
Copy link
Member

@ajdawson If you look at the travis output for the python 2.7 job you can see that running it in serial completed without a problem in just under 15 mins

@djkirkham djkirkham changed the title Bump nose process-timeout value Run Travis tests in parallel Aug 24, 2017
@djkirkham djkirkham changed the title Run Travis tests in parallel Run Travis tests in serial Aug 24, 2017
@pp-mo pp-mo merged commit 520640f into SciTools:dask_mask_array Aug 24, 2017
@pp-mo
Copy link
Member

pp-mo commented Aug 24, 2017

What a relief !!!

@ajdawson
Copy link
Member

There may be a happy medium somewhere. I've also never seen timeouts locally, but I've never used more than 8 cores for testing. I think Travis nodes present 64 cores which our tests will use. How do things look on 4 cores say?

@lbdreyer
Copy link
Member

@djkirkham can you make an issue to remind us to revert this to ensure it is only temporary?

@lbdreyer
Copy link
Member

lbdreyer commented Aug 24, 2017

I tested it with different numbers of processors and got the following results (note that I only ran each test once so the timings won't be that reliable):

No. of processors py27 default tests py27 min test py27 coding py27 example py27 doctests py35 default tests py35 minimum test py35 coding tests py35 example tests py35 doctests
1 15m11 5m41 6m16 8m11 9m22 11m55 4m13 8m10 7m23 10m24
4 6m56 5m29 6m3 6m57 10m8 6m 6m12 6m56 7m4 10m17
8 5m23 5m31 6m32 6m19 6m37 4m26 3m47 6m9 6m34 9m29
12 11m253 5m36 6m29 7m2 10m1 8m22 6m23 6m28 6m53 15m57
16 TimedOut 3m59 4m19 6m40 10m25 TimedOut 5m49 6m47 7m28 9m51

I could do more testing but I think setting 8 processors might be our best bet

@QuLogic QuLogic added this to the dask-mask milestone Aug 24, 2017
@djkirkham djkirkham deleted the bump_nose_timeout branch October 26, 2017 13:00
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants