Closed mathieucarbou closed 2 years ago
FYI:
First build after PR was complete passed successful - linux builds on azure lasted ~40 min (10 min less than before).
Now that the build is stable, I added a commit to try again with fork count 4 on Azure this time. If the build duration considerably decreases without impacting the test reliability, we will be able to keep fork count 4.
FYI:
First build after PR was complete passed successful - linux builds on azure lasted ~40 min (10 min less than before).
Now that the build is stable, I added a commit to try again with fork count 4 on Azure this time. If the build duration considerably decreases without impacting the test reliability, we will be able to keep fork count 4.
Rolling back: forkcount 4 slows down the container too much.
Best numbers are with forCount 3: with forCount 3, all 4 builds always passed, 2 times consecutively, for now out of 2 runs.
I'm currently preparing an EE PR.
Let's wait for the new release of terracotta-utilities first also, that I will include in both angela and platform.
Please look at the commit list.
This PR contains several things:
Angela update: fixed several issues and potentially blocking call when closing agents (not impacting platform or EE but only projects using Ignote)
Setting default timeout to 30 seconds for tools
Extracted timeout parameters for commands dealing with node restart in a common class
Bugfixes
verbose mode was not working in test mode
correctly set entityOperationTimeout value according to help
correctly implemented stop and restart detection in config to remove any sleep or impact due to connection timeout
Tests
Add DcActiveVoter to allow listening for Voter votes instead of doing an active sleep => also display voter errors and timeouts
Fixing test timeouts:
waitFor...
Results:
This is more than 15-20min decrease in test duration when run sequentially. Running with 4 forks locally, but just 3 on Azure. On azure, with fork count 3, build duration is reduced by more than 10 min.
main branch:
PR #1082 in platform