Most of the recent pull request checks in Keycloak fail because some of the Travis builds time out after 50 minutes. This started roughly a week ago (around Aug 1, 2019). There was no change in the code base that would cause this long test runs, so it seems that infrastructure is causing this. Is this any known issue?
Yes! Last 5 days we have a network issues. They are totaly random. From failing apt-get install command to timeouted docker push. I have already created a ticket on these problems yestarday, but with not response from Travis
Thank you for the suggestion. We are already working on that part as well, but I think this is not the core of the issue, since that happens just the same for the jobs that were finished in about 25 minutes. In some cases, the first - compilation - step failed after 50 minutes while usually it finishes within 10 minutes.
Very similar case for us (and there are several other threads opened recently for network timeouts in various situations):
Get https://registry-1.docker.io/v2/: net/http: request canceled while waiting for connection (Client.Timeout exceeded while awaiting headers)
+rm -rf -- /home/travis/build/apache/airflow/.build/cache/UVy7yccSkc
If you enable some output on the “compiling…” stage, we’d get something to work with to find out what takes it so long.
It could be the network issues. Or it could be e.g. excessive thrashing due to too many jobs running in parallel. Java build jobs are known to use up to multiple gigabytes of memory each, and a build machine has 7.5GB.
