Postgres Cluster (Restore) is able to be created with name that causes pod to have label longer than 63 characters.

[ncrmro]

Overview

Create cluster with long name unsupervised-shared-usecase-postgres-restore. In our case we where creating a restore. The restore completes fine but non of the stateful sets are able to

 Warning  FailedCreate  21m (x66 over 13h)     statefulset-controller  create Pod unsupervised-shared-usecase-postgres-restore-primary-78cd-0 in StatefulSet unsupervised-shared-usecase-postgres-restore-primary-78cd f │
│ ailed error: Pod "unsupervised-shared-usecase-postgres-restore-primary-78cd-0" is invalid: metadata.labels: Invalid value: "unsupervised-shared-usecase-postgres-restore-primary-78cd-5979c86d57": must be no more than 6 │
│ 3 characters    

Environment

Please provide the following details:

• Platform: AKS v1.27.8
• Platform Version: (e.g. 1.20.3, 4.7.0)
• PGO Image Tag: ubi8-15.4-0
• Postgres Version: 15
• Storage: EBS

Steps to Reproduce

REPRO

Create cluster with long name unsupervised-shared-usecase-postgres-restore. In our case we where creating a restore. The restore completes fine but non of the stateful sets are able to

 Warning  FailedCreate  21m (x66 over 13h)     statefulset-controller  create Pod unsupervised-shared-usecase-postgres-restore-primary-78cd-0 in StatefulSet unsupervised-shared-usecase-postgres-restore-primary-78cd f │
│ ailed error: Pod "unsupervised-shared-usecase-postgres-restore-primary-78cd-0" is invalid: metadata.labels: Invalid value: "unsupervised-shared-usecase-postgres-restore-primary-78cd-5979c86d57": must be no more than 6 │
│ 3 characters    

EXPECTED

Cluster to restore and be running with restored dataset.

ACTUAL

Cluster restore finished and stateful sets could not create their pods because label is to long.

[tjmoore4]

Hi @ncrmro. Thank you for reaching out; I'm sorry to hear you're hitting this issue. Fundamentally, there is an inherent limit to the maximum length for certain objects due to limitations in Kubernetes itself (please see the docs). With that said, I've added some notes and referenced your issue in our backlog to further clarify our documentation in this regard.

[ncrmro]

@tjmoore4 my though here is the operator should not allow a cluster name that will fail midway through a restore

[tjmoore4]

@ncrmro Understood. I've added an additional issue to track this request and evaluate potential improvements. Thanks again for reaching out.

[ncrmro]

@tjmoore4 thank you, yall are doing amazing work! Feel free to close if needed!