From 8e712e49fbdb99d39b2687dbf69b8f0ca220a91b Mon Sep 17 00:00:00 2001
From: Rafi Shamim <rafi@cockroachlabs.com>
Date: Thu, 12 Dec 2024 00:10:40 -0500
Subject: [PATCH] jobs: don't retry revert if permanent job error occurs

Release note: wip
---
 pkg/jobs/registry.go | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/pkg/jobs/registry.go b/pkg/jobs/registry.go
index fffa4f734822..796b47f9487e 100644
--- a/pkg/jobs/registry.go
+++ b/pkg/jobs/registry.go
@@ -1758,6 +1758,10 @@ func (r *Registry) stepThroughStateMachine(
 			// mark the job as failed because it can be resumed by another node.
 			return errors.Errorf("job %d: node liveness error: restarting in background", job.ID())
 		}
+		if IsPermanentJobError(err) {
+			// If there was a permanent error while reverting, then give up on reverting.
+			return r.stepThroughStateMachine(ctx, execCtx, resumer, job, StatusRevertFailed, err)
+		}
 		return onExecutionFailed(err)
 	case StatusFailed:
 		if jobErr == nil {
@@ -1775,9 +1779,6 @@ func (r *Registry) stepThroughStateMachine(
 		r.removeFromWaitingSets(job.ID())
 		return jobErr
 	case StatusRevertFailed:
-		// TODO(sajjad): Remove StatusRevertFailed and related code in other places in v22.1.
-		// v21.2 modified all reverting jobs to retry instead of go to revert-failed. Therefore,
-		// revert-failed state is not reachable after 21.2.
 		if jobErr == nil {
 			return errors.AssertionFailedf("job %d: has StatusRevertFailed but no error was provided",
 				job.ID())