[schwarztrinker]

Refs #1043

What changed

scripts/verify-fresh-node.sh, phase_delete only:

• header comment: the D18 contract replaces the "operator has no delete path yet" note;
• new step 1 deletes lakehouse/${LAKEHOUSE} first (its finalizer uninstalls trino/lakekeeper/openfga/seaweedfs and retains the catalog PVC); a failed Lakehouse delete exits 1;
• step 2 is the PlatformInstance delete (was step 1);
• new wait loop kubectl wait --for=delete deploy/{nova,keycloak} --timeout=60s before the Deployment assertions, with || true so the get loop stays the assertion;
• the remaining step numbers shift by one (3, 4, 5) — the ticket only names the 1→2 renumber, this keeps the sequence consistent.

Checks

$ bash -n scripts/verify-fresh-node.sh && echo "syntax ok"
syntax ok
$ git diff --stat
 scripts/verify-fresh-node.sh | 31 ++++++++++++++++++++++++-------
 1 file changed, 24 insertions(+), 7 deletions(-)
$ ./scripts/verify-fresh-node.sh; echo "exit=$?"        # readiness phase on the lab, read-only
? instance "platforminstance/cluster is not readable in stardelt"
verify-fresh-node: UNVERIFIED (no PlatformInstance on this cluster)
exit=3
$ make lint
lint: OK

Function-level harness (stub kubectl, phase_delete extracted verbatim, scratch under $HOME), three scenarios:

=== A: full success ===   exit=0
kubectl -n stardelt delete lakehouse/main --wait=true --timeout=30m
kubectl -n stardelt delete platforminstance/cluster --wait=true --timeout=30m
kubectl -n stardelt wait --for=delete deploy/nova --timeout=60s
kubectl -n stardelt wait --for=delete deploy/keycloak --timeout=60s
kubectl -n stardelt get deploy/nova           # the assertion after the wait
=== B: Lakehouse delete fails === exit=1
✗ delete "lakehouse/main was not deleted"
verify-fresh-node: FAIL (the Lakehouse delete did not complete)
  (log stops there: the PlatformInstance delete is never issued)
=== C: GC never removes nova/keycloak === exit=1
✗ delete nova still present
✗ delete keycloak still present
verify-fresh-node: FAIL (the delete left something behind)

Not run

The real delete phase on a fresh node (destructive, owned by platform#293 on demand; the lab has no PlatformInstance). That run is the final proof, after operator#587/#588 are in the pin (operator#587 is already there, platform#1046).

Remaining

Nothing in this ticket.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota.

An organization owner can choose the repository organization as the billing source for reviews requested by members with a paid Copilot license. Learn about Copilot code review billing.