Compute worker - docker pull image + execution time limit crash

[ihsaan-ullah]

@ mention of reviewers

@Didayolo

A brief description of the purpose of the changes contained in this PR.

Now users can see the following error logs:

1. docker image pull failure logs in ingestion std_err
2. execution time limit logs in ingestion or scoring std_err

Screenshots

Docker image pull

Execution time limit Ingestion

Execution time limit Scoring

Issues this PR resolves

• No error logs when the compute worker crashes #902

Checklist

• Code review by me
• Hand tested by me
• I'm proud of my work
• Code review by reviewer
• Hand tested by reviewer
• CircleCi tests are passing
• Ready to merge

[Didayolo]

That is awesome!

In case of another problem (not docker pull or time limit), is the error log transmitted too?

[ihsaan-ullah]

That is awesome!

In case of another problem (not docker pull or time limit), is the error log transmitted too?

I cannot think of any other problem which may occur outside the running of the submission. Docker pull is happening before running the submission and execution time limit is also outside the submission execution. But if there is any other scenario, I can check that

[Didayolo]

I think that was the main scenarios indeed. Others scenarios are basically bugs inside the code of the compute worker. For instance de BadZipFile that we have fixed already. We can just assume there is no other bug while nothing is reported.

[ihsaan-ullah]

I think that was the main scenarios indeed. Others scenarios are basically bugs inside the code of the compute worker. For instance de BadZipFile that we have fixed already. We can just assume there is no other bug while nothing is reported.

Yes, I agree. Now I think if there is any other bug, it can be fixed in less time as now things seems to be working

[Didayolo]

I suggest to test this PR in this way:

• Build and push the compute worker docker to codalab/competitions-v2-compute-worker:test (note the test tag).
• By the way, we can remove the image with tag retries from DockerHub.
• Update the compute workers of the default queue of https://codabench-test.lri.fr/ with this new docker image.
• Try different submission on codabench-test, exceeding execution time limit or not, etc.

This procedure can be used for future developments too.

[ihsaan-ullah]

Push the compute worker docker to codalab/competitions-v2-compute-worker:test (note the test tag).

why the compute worker docker image.

[Didayolo]

why the compute worker docker image.

To reflect changes done on compute_worker.py, we need to build the docker image and push it to dockerhub.

[Didayolo]

@ihsaan-ullah

I updated the compute worker associated to codabench-test server with your changes.

We need to make submission to test this PR.

https://codabench-test.lri.fr/

Seems to be working fine!