fix(ops): fail rebuild fast when the referenced backup is missing

[weicao]

Problem

A RebuildInstance OpsRequest using the in-place path (rebuildFrom[].inPlace: true) with a backupName that references a nonexistent Backup never fails — the ops controller requeues and retries forever, and the OpsRequest stays Running indefinitely. The same applies when the ActionSet referenced by backup.status.backupMethod.actionSetName has been deleted.

Both conditions are deterministic and permanent: validateRebuildInstance does not validate the referenced backup at submit time, and a deleted Backup/ActionSet does not come back.

Root cause

In pkg/operations/rebuild_instance.go, prepareInplaceRebuildHelper:

• the cli.Get of the Backup named by rebuildFrom[].backupName returned the bare NotFound error;
• dputils.GetActionSetByName for backup.status.backupMethod.actionSetName also returned the bare NotFound error.

In the operations framework, a plain error from ReconcileAction means "transient, requeue"; only intctrlutil.NewFatalError is treated as terminal (rebuildInstancesInPlace catches ErrorTypeFatal, marks the instance progress Failed, and lets the OpsRequest transition to Failed). The adjacent validations in the same function (backup not Full/Incremental, backup not Completed, empty backupMethod) already return NewFatalError; only these two NotFound lookups leaked through as retryable.

Change

• Classify NotFound on the Backup lookup as intctrlutil.NewFatalError, matching the neighboring checks.
• Classify NotFound on the ActionSet lookup (same function, same error-flow) as intctrlutil.NewFatalError.
• Add two regression tests driving ReconcileAction end to end: missing backup and missing actionSet both mark the instances' progress Failed with a "not found" message and move the OpsRequest to Failed on the next reconciliation.

Not changed

• Non-NotFound errors from both lookups still return as-is (retryable), preserving behavior for transient API errors.
• The hscaling path, the Action phase instance lookup fixed by fix(ops): fail rebuild ops for non-existent instances #10510, and submit-time validation are untouched.

RED evidence

Both new tests fail on origin/main (before the fix) with the bare NotFound error escaping ReconcileAction:

[FAIL] ... [It] test rebuild instance in place when the backup does not exist
[FAIL] ... [It] test rebuild instance in place when the actionSet of the backup does not exist
    Unexpected error: backups.dataprotection.kubeblocks.io "backup-not-exist-..." not found
    Unexpected error: actionsets.dataprotection.kubeblocks.io "actionset-not-exist-..." not found
Ran 2 of 114 Specs ... 0 Passed | 2 Failed

With the fix applied, the focused run and the full go test ./pkg/operations/... -count=1 suite pass; go vet and gofmt -l are clean.

Fixes #10542

🤖 Generated with Claude Code

[apecloud-bot]

Auto Cherry-pick Instructions

Usage:
  - /nopick: Not auto cherry-pick when PR merged.
  - /pick: release-x.x [release-x.x]: Auto cherry-pick to the specified branch when PR merged.

Example:
  - /nopick
  - /pick release-1.1

CLA Recheck Instructions

Usage:
  - /recheck-cla: Trigger a re-check of CLA status for this pull request.
Example:
  - /recheck-cla

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 63.67%. Comparing base (ecd604d) to head (d8c83ea).
⚠️ Report is 15 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #10544      +/-   ##
==========================================
- Coverage   63.68%   63.67%   -0.01%     
==========================================
  Files         519      519              
  Lines       62852    62854       +2     
==========================================
- Hits        40027    40023       -4     
- Misses      19202    19211       +9     
+ Partials     3623     3620       -3     

Flag	Coverage Δ
unittests	63.67% <100.00%> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[leon-ape]

[P2, non-blocking] The new regression test consumes an ActionSet before its owner has published a valid current-generation result.

The test creates an ActionSet with no status.phase or status.observedGeneration and immediately expects rebuild to proceed (pkg/operations/rebuild_instance_test.go:932). ActionSet exposes those status fields specifically so its controller can report whether the current spec is valid. Please make the fixture current-generation Available before asserting that rebuild consumes it; otherwise the test codifies the existing implementation detail that object existence alone is sufficient.

The previous production issue where ActionSet NotFound was made terminal has been correctly reverted on this head; there is no remaining production blocker from that feedback.

[weicao]

Handled the non-blocking fixture gap in exact head 4687616203fcff6c214ca002e6db3e58393dc76a. The late-installed ActionSet is now patched to status.phase=Available with status.observedGeneration == metadata.generation before the regression expects rebuild consumption. No production behavior was added; the prior ActionSet-NotFound fatal path remains reverted. Focused regression, full pkg/operations, go vet ./pkg/operations, and diff-check pass locally. The PR remains Draft for exact-head CI/review verification.