Smartlyio/fetch latest runner prerelease

[sjagoe]

Identify the Bug

#74: Runners don't pick up the first job that triggered scale-up due to updating to a pre-release of github actions runner.

Description of the Change

This change causes the runner sync lambda to pre-cache the latest pre-release of the github actions runner to prevent auto-update on start.

It achieves this by listing all releases, which includes pre-releases, and selecting the first release, which should be the most recent.

Alternate Designs

N/A

Possible Drawbacks

It is not explicitly documented that releases are provided in the order expected by this change. If the order provided by github changes, this will fail unless all pages of releases are retrieved and sorted by the client.

Verification Process

• The unit tests are updated to match the expected API.
• The code included here has been running live for about a week without any issues.

Release Notes

• syncer will now sync the latest github actions runner pre-release, preventing any auto-updating on startup

[npalm]

Really great the work you did, but I having a second thought. I assume you had the same problem as we have. At GitHub they have crearted a pre-release. and did not make the release final. Which is in my opnion a mistake. Or otherwise the update script of the runner is trying to update with a pre release. Which I would call a bug. (see actions/runner#581)

As you mentioned the order of releases is not documented, but it seems the config.sh scripts uses also the listReleases api. I still hope to get a answer from GitHub what they think should be the approach. @gertjanmaas any opinion?

[gertjanmaas]

I agree with @npalm

[sjagoe]

I agree; I'd really like to see this fixed by github; it shouldn't be up to this module to work around the pre-releases.

In any case, I'll continue to maintain the branch in my fork of your repo and use it internally, because we saw the pre-release update issues occur again only two weeks ago (see https://github.com/philips-labs/terraform-aws-github-runner/issues/74#issuecomment-669074264), and that pre-release was left in that state for at least a week; during that time we had jobs not running in our AWS pool.

I see that now we have runner v2.273.0 in pre-release. I would expect jobs to fail to run again due to the update on start.

[gertjanmaas]

Based on this PR and the trouble with the latest prerelease, I've created #165 to make this an option. Does this suit your needs @sjagoe ? If so I propose to close this PR in favour of #165

[sjagoe]

@gertjanmaas Thanks, #165 looks great. I'll go ahead and close this.