#7041 fix(propeller): treat K8s BadRequest/Invalid as permanent failure instead of retrying Pull request opened by themavik Summary • Uncomment the

PhaseInfoFailure

return for `BadRequest`/`Invalid` K8s errors in

plugin_manager.go

so they are treated as permanent failures instead of being retried indefinitely Problem When a Kubernetes API server returns a

BadRequest

(400) or

Invalid

(422) error during resource creation (e.g., an admission webhook rejection), FlytePropeller logs the error but falls through to the generic system error handler. This causes the task to be retried indefinitely, wasting resources and delaying user feedback. The return statement on line 264 was commented out: } else if k8serrors.IsBadRequest(err) || k8serrors.IsInvalid(err) { logger.Errorf(ctx, "Badly formatted resource for plugin [%s], err %s", e.id, err) // return pluginsCore.DoTransition(pluginsCore.PhaseInfoFailure("BadTaskFormat", err.Error(), nil)), nil } Without the return, execution falls through to line 271 which wraps the error as a generic system error and returns

UnknownTransition

, causing infinite retries. Root Cause The sibling code in

flyteplugins/go/tasks/plugins/array/k8s/subtask.go

handles the same case correctly (uncommented), confirming this was an oversight. Fix Uncomment the return statement so `BadRequest`/`Invalid` errors immediately transition to

PhasePermanentFailure

with a clear

"BadTaskFormat"

reason. Added a unit test (

jobBadRequest

) that verifies the fix. Testing • Added

TestK8sTaskExecutor_Handle_LaunchResource/jobBadRequest

test that creates a fake client returning

k8serrors.NewBadRequest(...)

and verifies the transition phase is

PhasePermanentFailure

• Existing tests remain unchanged Closes #6531 flyteorg/flyte