1543490 - Retain *two* latest generic-worker task directories in Remove-OldTaskDirectories

Status: RESOLVED FIXED

(Infrastructure & Operations :: RelOps: OpenCloudConfig, task)

Description

[Pete Moore [:pmoore][:pete]]

After upgrading generic-worker to release 14.0.2 in production today, tasks using sccache failed, and we had to roll back the upgrade.

The new release of generic-worker contained a fix for bug 1533694 which meant that both a task folder for the current task, in addition to a task folder for the task to be run after the next reboot, existed at the same time on the file system.

OpenCloudConfig currently has a workaround in place for the generic-worker defect described in bug 1433854, to routinely purge all but the latest task folder from the file system. This works for releases prior to generic-worker 14.0.2 which never had two active task folders on the file system at the same time. However, with generic-worker 14.0.2, this purging logic needs some slight adjustment.

So although generic-worker prepared the task folder for the task that should run after the following reboot, OpenCloudConfig was deleting it, and this is what broke sccache.

The fix for this bug is to get OpenCloudConfig to retain the latest two task folders (rather than just the latest task folder). In addition, I intend to make a generic-worker for bug 1433854 in the coming days, so that we won't need Remove-OldTaskDirectories any longer in OCC (which is the function that purges the old task folders).

Comment 1

[Pete Moore [:pmoore][:pete]]

Attachment: GitHub Pull Request for OpenCloudConfig