Batch Job Garbage Collection #3982
Conversation
| if remaining <= available { | ||
| for _, job := range jobs[submittedJobs:] { | ||
| jns := structs.NamespacedID{ID: job.ID, Namespace: job.Namespace} | ||
| req.Jobs[jns] = option |
There was a problem hiding this comment.
Little nervous about sharing a pointer to the same struct here, but nothing mutates it and it seems unlikely anything would ever go back through and mutate it so I think it's safe to leave.
| return err | ||
| } | ||
|
|
||
| // If the job is periodic or parameterized, we don't create an eval. |
There was a problem hiding this comment.
The scheduler doesn't do anything for those job types. It does work for the derived batch jobs from them
There was a problem hiding this comment.
Ah sure, that makes sense. Maybe in future comments add a because ... clause as what the code is doing is fairly obvious, but why it's doing it is not.
nomad/job_endpoint.go
Outdated
| } | ||
|
|
||
| // Lookup the job | ||
| snap, err := j.srv.fsm.State().Snapshot() |
There was a problem hiding this comment.
Do this once outside the loop
|
I'm going to lock this pull request because it has been closed for 120 days ⏳. This helps our maintainers find and focus on the active contributions. |
This PR introduces a new endpoint to batch deregister jobs. This allows the
garbage collector to remove many jobs more efficiently.