xchwan commented on PR #75945:
URL: https://github.com/apache/airflow/pull/75945#issuecomment-6898700238

    @stegololz  Thanks for the review. Summary of what this update changes and 
how it was tested.
   
     **What changed**
     - `dagProcessor.sets` names are validated: `default` is reserved, names 
must be valid DNS-1123 labels, and duplicates are rejected.
     - ServiceAccounts of the sets are bound once each in the SCC role binding.
     - Fixes for the PDB overlap, ServiceAccount handling and custom `args` 
with sets.
     - `NOTES.txt` warns when a Dag bundle is selected by more than one set.
   
     **What was tested**
     - Helm unit tests (`test_dag_processor_sets.py`, 
`test_scc_rolebinding.py`): 135 passed.
     - Real install on kind (k8s v1.31.12, LocalExecutor) with 3 bundles and 
two sets:
       - one Deployment per set and no default one, with the expected 
`--bundle-name` args; each processor only parsed its own bundles
       - correct ServiceAccounts (including a custom one), PDBs per set, 
`rollout restart` works
       - `enableDefault=true` next to sets adds the default processor and its 
PDB
       - invalid set names (`default`, `Team_A`) fail at render, before 
anything is created
       - `/api/v2/dags` returns the right `bundle_name` per Dag; all three Dags 
ran to success
     - Upgrade without `sets`: the chart from `upstream-main` and this branch 
render identical Dag processor Deployment, ServiceAccount, PDB
     and SCC role binding for the default setup, a custom ServiceAccount + PDB, 
custom `replicas`/`args`, and SCC enabled.
   
     **Help needed: OpenShift SCC**
     I could not test this on a real OpenShift cluster (kind has no SCC 
admission). What is verified: with `createSCCRoleBinding=true`, the
     role binding lists each ServiceAccount (including the sets' ones) exactly 
once, all of them exist in the namespace, and the unit tests
     cover the `enableDefault`/custom ServiceAccount combinations.
     Could someone with access to an OpenShift cluster check that the Dag 
processor pods of the sets are admitted and start under the `anyuid`
     SCC?
   
     ---
     Generated-by: Claude Code (Sonnet 5.5); 


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to