Build is not triggered on each commit(after hours) of PR

[ironakpatel]

we are 1.4.0-RC2, and noticing any new commit(after hours/late nights EST) to open PR is not triggering pipeline. is there any config that needs to be updated or is this known issue/feature?

[esselius]

My company have noticed this issue too, no errors in logs, but I could manually trigger the pipeline with options, finding the commit :s

[dlethin]

We are also seeing issues with some commits to PRs not triggering pipeline runs. While the original submitter mentioned this happens after hours, I haven't gone that far so as to establish any pattern yet why some work and some don't. We just have the same symptoms -- no trigger, no errors in the log files. We're using github-pr-poller-1.3.4.jar and currently on GoCD Server 19.6.0

[esselius]

@ironakpatel @dlethin can you reproduce this: #71 ?

[dlethin]

@esselius - I'm a bit of middle-man here, helping support a team with GoCD that reported this problem. While I'm told this has happened multiple times, I only researched one instance, but I can confirm that was indeed a merge commit. I'll have to do some digging sometime next week and see if that's the pattern for others as it sounds plausible.

As far as the workaround you mentioned on running the plugin with options and selecting the commit, I'm familiar with finding commits for branches, but that works for PRs too? Is there an easy way to find that commit hash in the UI when looking at the PR ? Or are you saying to you take the commit hash from the branch that's being merged? Thanks.

[esselius]

@dlethin I just pressed the Trigger with Options-button and selected the top commit, which happened to be the merge commit :)

[philip-hope]

We're seeing the same symptoms in our setup sporadically, using v1.3.4: delayed runs across multiple repos up to several hours in some instances for both default and merge commits
So far we aren't seeing any explicit errors in the logs either nor a definitive pattern as to what is causing it, and are investigating the possibility of network performance issues between GoCD and Github, as well as where in the source code an instance may be hung or queued.

[philip-hope]

@ashwanthkumar Do you have any more context on why this might be happening? Despite any resolutions in #71 #83 and #94 we're still experiencing seemingly random delays each few days to the pr builder triggering for default and merge commits, causing delays to multiple users across our platform.

For reference we're using GoCD v20.2.0 & pr builder: v1.3.4

Having monitored this for a few weeks now we're not able to locate any specific errors. We have however noticed the following in the server logs:

Skipping update of material 'PluggableSCMMaterial{[url=https://[github_repository]]}' which has been in-progress since Tue Apr 13 18:21:47 BST 2021
The highest recorded in-progress events have been up to several hours long, though most often delays span an hour.

Monitoring our GoCD health-check endpoint api/v1/health, we understood a delay over 15 minutes would affect this status however no changes were recorded.

The solution to trigger the pr-builder with the specific commit (gocd trigger with options) proves successful if the commit is present, however there have been several occurrences when the pipeline is delayed and the latest commit is not known/available in GoCD.

We have additionally tested the behaviour of merge commits vs rebase commits in the hope that merge commits are consistently delayed, however both triggered as expected.

Any guidance you can offer is much appreciated!