More robust handling when Sentry has HTTP 5xx error

[iBotPeaches]

Environment

Standard GitHub Runners

Version

v3.4.0

Steps to Reproduce

1. Use Action
2. Use it during a specific time and place when Sentry has an issue

Sadly this treats our build as a failure and thus needs a re-run. Out of our last 30 build failures in past 22 days, 11 of those are Sentry having an HTTP 5xx error during this action's usage.

Expected Result

The Action to retry on 5xx with exponential backoff?

Actual Result

Error: Command failed: /action-release/node_modules/@sentry/cli-linux-x64/bin/sentry-cli --header sentry-trace:xxx-1 --header baggage:sentry-environment=production-sentry-github-action,sentry-release=3.4.0,sentry-public_key=xxx,sentry-trace_id=xxx,sentry-sample_rate=1,sentry-transaction=sentry-github-action-execution,sentry-sampled=true releases set-commits v0.5.0-xxx --auto
error: API request failed

Caused by:
    sentry reported an error: gateway timeout (http status: 504)

Add --log-level=[info|debug] or export SENTRY_LOG_LEVEL=[info|debug] to see more output.
Please attach the full debug log to all bug reports.

node:internal/errors:984
  const err = new Error(message);

[linear]

GHA-27

[FGoessler]

Also seeing increased number of 504 failures that fail our builds. Retry logic AND stabilizing the API would be great. ;)

[szokeasaurusrex]

Hi @iBotPeaches and @FGoessler, thank you for this suggestion; adding retry logic should be pretty simple to do!

These requests are being made by Sentry CLI, which this action wraps. We added more robust retry logic to Sentry CLI in version 2.46.0, but this action is pinning Sentry CLI 2.41.1.

So, we should be able to get that additional retry logic with a simple dependency upgrade 🙂

[andreiborza]

This has been released with 3.5.0.