Recovery: allow sbr as default, update docs and add missing flag to slot reset

[tkatila]

No description provided.

[Copilot]

🟡 Changes recommended

The webhook test mismatch and unresolved sample and documentation issues must be addressed before approval.

Get a fresh assessment by requesting another Copilot review.

Pull request overview

This PR enables SBR as a recovery default, updates slot-reset commands, and synchronizes documentation, schemas, samples, and tests.

Changes:

• Adds SBR to validation and CRD enums.
• Adds --force-reset-gpus to slot-reset commands.
• Updates recovery documentation, samples, and related tests.

File summaries

File	Summary
RECOVERY.md	Documents SBR and updated reset commands.
internal/controller/gpurecoveryplan_recoverytype_test.go	Tests SBR selection.
internal/controller/gpurecoveryplan_jobs_test.go	Verifies the updated slot command.
internal/controller/gpurecoveryplan_helpers.go	Updates fallback behavior and slot command generation.
config/samples/recoveryplan/gpurecoveryplan.yaml	Updates sample recovery configuration.
config/deployments/xpum/xpum-reset-job.yaml	Updates reset command documentation.
config/crd/bases/intel.com_gpurecoveryplans.yaml	Updates the generated CRD enum.
charts/gpu-base-operator/crds/gpurecoveryplans.yaml	Updates the Helm CRD schema.
api/v1alpha1/gpurecoveryplan_webhook.go	Accepts SBR as a default reset type.
api/v1alpha1/gpurecoveryplan_webhook_test.go	Tests accepted default reset types.
api/v1alpha1/gpurecoveryplan_types.go	Updates API validation markers and documentation.

Review details

Suppressed comments (5)

RECOVERY.md:21

• dependant is misspelled here; the adjective is dependent.

Recovery functionality is dependant on both the underlying system as well as the installed Intel GPU.

RECOVERY.md:25

• The new hardware table and surrounding text remove the previous caveats that BMG reset efficacy is still unvalidated and that the health-xpumd-gpu.wedged taint key is not confirmed against a shipping DRA driver. This PR only changes command selection and validation and adds no implementation or compatibility evidence for either caveat, so dropping them makes recovery support look more complete than it is; please retain them until the corresponding validation lands.

| SBR | A BMG Pro card (B50, B60 etc.). |

config/samples/recoveryplan/gpurecoveryplan.yaml:58

• This sample now points at registry.local/xpu-smi:devel, but no other repository file defines or documents that registry and the recovery guide still uses public image examples. Applying the checked-in sample in a normal cluster will make the recovery Job image unpullable; please keep the previous resolvable example or document this as a required substitution.

    image: "registry.local/xpu-smi:devel"

config/samples/recoveryplan/gpurecoveryplan.yaml:72

• This changes the firmware source to registry.local/intel-gpu-fw-binaries:devel, which is not defined or documented anywhere else in the repository, while RECOVERY.md still tells users to use the Docker Hub image. A user applying this sample will fail the firmware image preflight or Job pull unless they first know to rewrite the reference; retain the documented image or explicitly mark this as a required cluster-specific substitution.

        name: registry.local/intel-gpu-fw-binaries:devel

internal/controller/gpurecoveryplan_recoverytype_test.go:144

• The table entry says on a BMG Pro cards, which is grammatically incorrect; use on a BMG Pro card.

			Entry("sbr, on a BMG Pro cards", intelv1a1.RecoveryTypeSBR),

• Files reviewed: 11/11 changed files
• Comments generated: 4
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.