-
Bug
-
Resolution: Unresolved
-
Blocker
-
None
-
Jenkins 2.186 (from the jenkins/jenkins:2.186 docker image)
Amazon EC2 plugin (1.44.1)
After upgrading to 1.44.1, we are observing the return of long-standing queue locking as described in this PR (https://github.com/jenkinsci/ec2-plugin/pull/346).
It looks like the issue is reintroduced within the scope of this commit.
I think it's because since the `clock` is only instantiated within `readResolve`, and not in the constructor of `EC2RetentionStrategy`, the problem is re-introduced for workers that are being newly created (not for workers that were persisted from XML). The clock may not exist when the `check` method is invoked, which prevents the strategy's ability to to skip the `NodeProvisioner.update()` cycle triggered with `internalCheck`, if the last check happened < 1 minute ago. The solution is probably to instantiate the clock in both the constructor and `readResolve`.
Downgrading back down to version `1.42` where we patch in the code introduced in the original commit (https://github.com/jenkinsci/ec2-plugin/pull/346) fixes the issue and we no longer observe these frequently occurring queue locks.