Docker: Set RMQ consumer_timeout to undefined

[unkcpz]

Fixes https://aiida.discourse.group/t/workchain-except/152/36

[sphuber]

Wouldn't it perhaps be better to add the advanced.confgi approach instead, which allows us to simply set the timeout to undefined? I think the following should work

Suggested change

	# 1000 hours in milliseconds
	echo "consumer_timeout=3600000000" >> "${RMQ_ETC_DIR}/rabbitmq.conf"
	cat > "${RMQ_ETC_DIR}/advanced.config" <<EOF
	%% advanced.config
	[
	{rabbit, [
	{consumer_timeout, undefined}
	]}
	].
	EOF

[unkcpz]

Ah, cool, I didn't know this. Will give it a build locally and test.

[unkcpz]

Changed and add test checking it is set, but better to run a calculation to test it really activated.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (fae2a9c) 80.73% compared to head (01f9036) 80.73%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #6189      +/-   ##
==========================================
+ Coverage   80.73%   80.73%   +0.01%     
==========================================
  Files         548      548              
  Lines       40074    40074              
==========================================
+ Hits        32349    32350       +1     
+ Misses       7725     7724       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[unkcpz]

Hi @sphuber, in discourse you mentioned you can reproduce the issue correct? Can you try with ghcr.io/aiidateam/aiida-core-with-services:sha-558be45 from this PR and ghcr.io/aiidateam/aiida-core-with-services:edge to make sure this solves the problem?

[sphuber]

@unkcpz I baked this recipe locally so that I can run a container and try and debug why the build on Github actions was failing. I managed to build and start a container, but only as the normal user. When I try to -u root the container hangs during startup because it cannot start the services as root for some reason. Is there a way to become root inside the container though? I want to check rabbitmqctl environment to make sure the consumer_timeout is properly set to undefined. Note that verdi status actually runs fine in my locally baked image. Don't understand why it fails on GHA.

[unkcpz]

Yes, I found the same that -u root will stuck the initial processes. To root, running a container first and run docker exec -it -u 0 <container id> bash to get into the running container as root.

[unkcpz]

The failed test is on arm64 build, I can test it from my mackbook.

[unkcpz]

I tested locally on my M1 Mac, and all passed. The failed test is the timeout of _docker_service_wait, I roll back to the old time.sleep method to make sure the profile is ready before the test. The downside is the test takes a bit longer. If this makes the test pass, I will not combine the conftest polish changes in this PR.

[sphuber]

Yes, I found the same that -u root will stuck the initial processes. To root, running a container first and run docker exec -it -u 0 <container id> bash to get into the running container as root.

Excellent, that worked, thanks. I verified that rabbitmqctl environment returns

      {consumer_timeout,undefined},

so I am pretty sure it worked.