Add support to Tolerations & nodeSelect to the restore phase with velero/velero-restore-helper

[didier-segura]

Hi,

The issue : "I can't restore a backup from a FSB S3 bucket (fr-par-1) to another region (by example : fr-par-2)."

Let me explain more.

We have a cluster with 2 node group to handle legacy application with one pod / one pvc in RWO.

To control the pod repartition we have added a toleration like this, so legacy application without this key "multiaz" can't start in the other zone "fr-par-2" :

`tolerations:

• key: "multiaz"
  operator: "Equal"
  value: "true"
  effect: "NoSchedule"`

I can't find any element to setup this toleration when the velero start the restore process.

So the pod in the restore-wait phase can't start without this Toleration define.

I have added the "Toleration" to the NodeGroup by the way.

I have also added the JSON patch that apply correctly :

`version: v1
resourceModifierRules:

• conditions:
  groupResource: statefulsets.apps
  patches:
  • operation: add
    path: "/metadata/labels/changeZone"
    value: "fr-par-2"
  • operation: add
    path: "/spec/template/spec/nodeSelector"
    value: '{ "topology.kubernetes.io/zone": "fr-par-2" }'
  • operation: add
    path: "/spec/template/spec/tolerations"
    value: '[{"key": "multiaz", "operator": "Equal", "value": "true", "effect": "NoSchedule"}]'
• conditions:
  groupResource: persistentvolumeclaims
  patches:
  • operation: replace
    path: "/spec/storageClassName"
    value: "scw-bssd-retain-fr-par-2"
    `

The PVC are well setup in new zone "fr-par-2", but the pod restoration are in Pending state because they don't have the Toleration key "multiaz:true", and if I kill the pod in this restore stage, the new pod start correctly in fr-par-2 with an empty volume (which is "normal").

Is it possible to add this ability (tolerations) into the velero chart somewhere ? or in the ConfigMap logic ?

Best regards !

[jenting]

@qiuming-best do you know velero restore helper supports above scenario?