Segmentation fault in main branch Docker Unit Tests

[github-actions]

Motivation

In the main branch config tests, Docker exited with status 139:
https://github.com/ZcashFoundation/zebra/actions/runs/6737581570/job/18315670392#step:5:229

Which is a memory safety error in Docker (or perhaps Zebra):
https://www.airplane.dev/blog/sigsegv-segmentation-fault-linux-containers-exit-code-139

This could be a hardware fault, or a Docker bug.

Auto-Opened Issue

GitHub Actions workflow Docker Unit Tests #247 failed.

Event: push
Branch: main
Commit: a1e476ec618943ce24a5e17d5016c3300259215e

^{Created by jayqi/failed-build-issue-action}

[mpguerra]

@gustavovalverde we've had 2 of these failures in the last 2 weeks. Is this something you can look into?

[gustavovalverde]

@mpguerra yes, sure.

[teor2345]

I deleted this part of the ticket, because the bug is still happening after we launched Zebra correctly in the tests:

possibly related to stopping the container during startup (see PR #7900 for details).

[teor2345]

We're also seeing exit code 137 (out of memory or SIGTERM) in the same tests in PR #7923:
https://github.com/ZcashFoundation/zebra/actions/runs/6802937935/job/18497528956?pr=7923#step:5:277

There's nothing in the Zebra config or on the Rust side that could cause this issue, and all the other similar tests are passing.

So here are some possibilities:

• the config file mapping into the Docker image is causing a memory error because Zebra still has it open when the container is shut down
• using exec in the entrypoint script is applying the bash process memory limits to Zebra, but Zebra needs different limits
• Zebra is using more memory on startup after we put the parameters in the binary

We could try sleeping before docker stop, or removing exec from the entrypoint?

[teor2345]

Exit 137 (OOM?) happened again on the main branch:
https://github.com/ZcashFoundation/zebra/actions/runs/6804547600

[mpguerra]

I've increased the priority on this one as it seems to be happening fairly regularly

[teor2345]

Error 137 and 139 in two separate main branch runs, details in #8004

[teor2345]

PR #8088 did not fix this issue, it failed in the merge queue:
https://github.com/ZcashFoundation/zebra/actions/runs/7187957663/job/19576993550#step:5:261

[teor2345]

Failed in the merge queue in:
https://github.com/ZcashFoundation/zebra/actions/runs/7200413180/job/19614737572?pr=8103#step:5:261

I think we could just ignore this error for now, because it's not actually impacting our ability to test, and we can't reproduce it outside this environment.