Skip to main content

Checks to run

These are read-only checks; each prints a verdict and exits non-zero when it proves something is wrong. Run them after an upgrade and nightly. verify-pgdump restores into a scratch database and drops it; it is not read-only in the strictest sense.

After an upgrade​

make step-verify (the target the upgrade path runs) runs the three checks below:

CommandWhat it provesA failure means
make verify-trino-authTrino accepts authenticated sign-in.Authentication to Trino is broken.
make verify-catalog-permissionsThe live Catalog grants against the OPA bridge.The live Catalog grants do not match expectations.
make verify-backups-allEvery volume has a backup mechanism and a recent backup, or an accepted ticketed exception.A volume is missing a backup mechanism or has a backup older than the limit.

Nightly​

CommandWhat it provesA failure means
make verify-backups-allEvery volume has a backup mechanism and a recent backup, or an accepted ticketed exception.A volume is missing a backup mechanism or has a backup older than the limit.
make verify-imagesEvery pinned ghcr.io/stardelt/* image is readable with the cluster's pull credential.Some pinned image is not readable with the cluster's pull credential.
make verify-realm-adminThe realm-admin Secret still signs in.The realm-admin Secret can no longer sign in.
make verify-oidc-clientsEvery confidential client's secret still authenticates.One or more confidential client secrets no longer authenticate.
make verify-pgdump PG_DUMP=allThe pg_dump CronJobs produce restorable dumps.A dump cannot be produced or restored successfully.

When you change one thing​

SituationCommandWhat it proves
After a TLS or sign-in changemake verify-trino-tlsThe in-cluster Trino path is TLS-only and token-authenticated.
After a Catalog/role changemake verify-catalog-permissionsThe live Catalog grants against the OPA bridge.
After a Cluster changemake verify-backups, make verify-backups-nova, make verify-backups-identity, or make verify-backups-openfgaWAL archiving per database, with a real object in the bucket.

How to read a failure​

A ✗/FAIL line names the object and the reason (paste the real verify-images output block from troubleshooting.md's image section as the example):

✓ jupyter-singleuser ghcr.io/stardelt/jupyter-singleuser@sha256:c3d45ec4… readable
✓ stardelt-operator ghcr.io/stardelt/stardelt-operator@sha256:e7548851… readable
✗ superset ghcr.io/stardelt/superset@sha256:b9ced85d81a0c9efd2ca993f27bba5470e4815aad9d246b93e3ca5e62d26e8cd 404: the digest is gone from the registry (garbage-collected?)
✓ airflow ghcr.io/stardelt/airflow@sha256:2d9504d1… readable
✓ nova ghcr.io/stardelt/nova@sha256:8562b91b… readable
✓ spark ghcr.io/stardelt/spark@sha256:7f9c3cbc… readable
FAIL (1 of 6 images is not readable)

UNVERIFIED plus exit 3 means the check could not see its subject (bucket, Secret, kube context) and proved nothing — fix the access, do not treat it as a pass.

Exit 1 means the check ran and found a real problem — link to the matching anchor in ./troubleshooting where one exists (image pull), otherwise to the component page under ../services/.

What runs by itself​

make install/make upgrade run verify-backup-config and step-verify; everything else is manual. verify-images needs the cluster's pull credential, so it is never a CI job.