PDB selector values do not match pod labels

[joeygrasso]

Your environment

Chart Version: 1.16.0

Helm Version: v3.18.3

Kubernetes Version: 1.32.4-gke.1698000

What happened?

I enabled PDB using the following:

connect.pdb.enabled: true
connect.pdb.minAvailable: 1

The PodDisruptionBudget resource gets created, but the selector's labels do not match the running pod's labels, so the PDB is ignored.

>> kubectl describe pdb/connect
Name:             connect
Namespace:        default
Max unavailable:  1
Selector:         app.kubernetes.io/component=connect,app.kubernetes.io/instance=connect,app.kubernetes.io/name=connect
Status:
    Allowed disruptions:  0
    Current:              0
    Desired:              0
    Total:                0
Events:
  Type    Reason  Age                  From               Message
  ----    ------  ----                 ----               -------
  Normal  NoPods  10m (x3 over 3h56m)  controllermanager  No matching pods found

What did you expect to happen?

I expect a PDB to be created, and when reviewing the PDB I should see the status update to reflect the following:

>> kubectl describe pdb connect-test
Name:                  connect-test
Namespace:        default
Max unavailable:  1
Selector:         app=connect
Status:
    Allowed disruptions:  1
    Current:              2
    Desired:              1
    Total:                2
Events:                   <none>

Note the selector change and now the status is updated to reflect the matching pods.

Steps to reproduce

1. Enable PDB
2. Run Helm upgrade
3. Review PDB w/ kubectl get pdb connect note that there are no pods.

Notes & Logs

Happy to hop on a call and share more specific details privately.

[volodymyrZotov]

Hi @joeygrasso ! Thank you for reporting this issue.
Hope will be able to take a look closely into that soon!

[joeygrasso]

@volodymyrZotov, thank you. I am happy to help provide more information as well.

The other thing I did not point out in my original message is that the connect.pdb.minAvailable: 1 setting is being totally ignored. In my example output, the PDB is still configured with Max unavailable: 1

I added the missing labels with connect.podLabels and the PDB was detected and applied, but the policy is still incorrect because the minAvailable setting is not being set.

Missing labels:

• connect.podLabels.app.kubernetes.io/instance: connect
• connect.podLabels.app.kubernetes.io/name: connect