ci: run Flux-Parsl integration tests

[jameshcorbett]

This should probably wait for #51 to go in.

Changes the parsl install from the default pip version to master, and adds an integration dockerfile and test script for testing integrations---not sure if that's the best structure.

[jameshcorbett]

I don't understand why the CI failed so if anyone could explain I would appreciate it 👍

[dongahn]

I don't understand why the CI failed so if anyone could explain I would appreciate it 👍

Looks like something is returning 1.

[dongahn]

Could be intermittent. Let me restart.

[dongahn]

For GitHub action CI debugging, https://github.com/marketplace/actions/debugging-with-tmate might come in handy.

Probably adding the following step would give you a terminal into the container.

- name: Setup tmate session
      uses: mxschmitt/action-tmate@v3

[dongahn]

This might also be a question for @SteVwonder and @andre-merzky. Are we capturing the logs from the commands running inside our containers in our GitHub action infrastructure? This may come in handy when we need to debug -- beyond being able to say, hey non-zero return from the CMD.

[SteVwonder]

Are we capturing the logs from the commands running inside our containers in our GitHub action infrastructure?

Not sure why the logs aren't popping up in this check. For all of the other CI runs that I've seen, if the docker container (and the processes in it) write to stdout/stderr, then it gets propagated to the GH actions log. I'm tempted to say this is just a weird GH bug/failure for that particular run.

In #51 , I added a flux-specific script that runs to try and summarize any Flux-specific failures that occurred. We could do the same for other components too if we want. We just need to get input from the other components for how to do that for their testsuits.

[dongahn]

@andre-merzky: let us know if you find anything. If you don't, we will just have to wait until @SteVwonder comes back.

[dongahn]

The same here. This PR has been pending for awhile. I would say we discuss to see if we can unstick the current roadblock.

[dongahn]

@jameshcorbett will address Mikhail's comment above. @andre-merzky will look at CI issues above,

[jameshcorbett]

CI passed! I'm now squashing all the incremental commits down to one. The CI should pass again, or I'll lose my mind.

[dongahn]

Wow! heroic @jameshcorbett... you should tell me what the issue was later :-)

[jameshcorbett]

Still a little bit of a mystery what went wrong. The only thing I changed was that I removed the if [singleton tests fail] exit 1; else [run flux-parsl integration] line.

Does someone want to go ahead and merge this?

[dongahn]

Thanks @jameshcorbett !