Gitpod resources are wasted by running multiple prebuild processes for same commit

[shaal]

Bug description

Pushing a commit result in 3 (three!) separate prebuild processes for no apparent reason.

https://gitpod.io/t/phase2/tufts-core/f3835ce4-aa28-4ab6-975a-7dc95d129e83
https://gitpod.io/t/phase2/tufts-core/f5862986-03e7-4268-87ea-fcaac8236931
https://gitpod.io/t/phase2/tufts-core/552b3fc5-c58f-4d23-ac52-af2ea71fd9b6

Steps to reproduce

1. Add a repo to team/project, with gitpod.yml setup for Github prebuild.
2. Push a commit
3. Check the prebuild pages and see that multiple prebuilds are running (usually 2)
   https://gitpod.io/t/team/project/prebuilds

Workspace affected

No response

Expected behavior

No response

Example repository

No response

Anything else?

No response

[AlexTugarev]

Thanks for reporting @shaal!

This is not what we expect to happen. I did a quick look at the webhook event registered, and see 3 distinct events:

|creationTime            |type        |authorizedUserId                    |rawEvent             |
|------------------------|------------|------------------------------------|---------------------|
|2022-10-27T13:46:47.363Z|pull_request|A                                   |                     |
|2022-10-27T13:46:45.764Z|push        |A                                   |w/ installation id   |
|2022-10-27T13:46:45.592Z|push        |B                                   |w/o installation id  |

The pull_request event is slipped through despite #13295, which is reasonable if different server pods handled the requests.

The new issue here is, the two push events look like they've been produced by two webhooks. While one is from the GH App, where is origin of the second event?

@shaal, do you have a change to inspect the Webhooks page for the phase2/tufts-core repository?

[AlexTugarev]

Following the current trend, a holistic way to solve this issue would be to move triggering of prebuilds into a control loop which would query the DB for recently recorded webhook events.

Also pull_request events should not trigger a prebuild at all. To clarify, there are two types of PRs which would be registered in those events. If the source repository is the same as the current target, a push event would also be registered. If the source is a different repository, we'd likely expect Gitpod would be configured there as well.

[shaal]

@AlexTugarev I noticed this is still happening

Where should I look?
The example in the screenshot is from a branch with no PR.
I pushed code to Github, and I see that 2 prebuilds are running in parallel.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[shaal]

This is still happening

[Konard]

I have the same issue again, please reopen the issue.