Job revert #2575
Job revert #2575
Conversation
nomad/job_endpoint.go
Outdated
| return fmt.Errorf("job %q not found", args.JobID) | ||
| } | ||
|
|
||
| if args.JobVersion == cur.Version { |
There was a problem hiding this comment.
Make sure JobByID's docs explain it returns the "latest" or "current" version of a Job ID now that an ID doesn't uniquely identify exactly one *structs.Job.
|
|
||
| // If the request is enforcing the existing version do a check. | ||
| if args.EnforcePriorVersion != nil { | ||
| if cur.Version != *args.EnforcePriorVersion { |
There was a problem hiding this comment.
This is racy, and probably should be evaluated inside the FSM. Effectively this is a check-and-set, but multiple concurrent calls to this API will be allowed to make progress depending on the timing.
There was a problem hiding this comment.
Yeah I realized that when implementing this. The register endpoint should push the logic from the RPC layer down into the FSM. This will fix this issue for both this and the register. Just out of scope for this PR. I will put a comment on that part of the code though
nomad/job_endpoint.go
Outdated
| return fmt.Errorf("job %q at version %d not found", args.JobID, args.JobVersion) | ||
| } | ||
|
|
||
| cur, err := snap.JobByID(ws, args.JobID) |
There was a problem hiding this comment.
I would check this first, since you expect to get the "job not found" error before the "job at version" not found error.
| } | ||
|
|
||
| reg.EnforceIndex = true | ||
| reg.JobModifyIndex = cur.JobModifyIndex |
There was a problem hiding this comment.
The modify index will get overwritten at Upsert time right?
There was a problem hiding this comment.
Those two are the way to do the CAS operation on the register
|
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 revert endpoint that allows a job to be reverted to a prior state.