Skip to content

Commit 591e87d

Browse files
committed
test(clickhouse): name what really leaves the round-trip's Secret behind
The suites run one at a time in name order, so in a CI run the projected Secret the preconditions delete comes from clickhouse-2-backup-roundtrip in the same namespace, not from the mongodb and rabbitmq suites that run later. The restore budget is also not spent on a credentials hold: by then the stamp is older than the backup's own two-minute wait. Assisted-by: LLM Signed-off-by: Alexey Artamonov <[email protected]>
1 parent e41688d commit 591e87d

1 file changed

Lines changed: 6 additions & 4 deletions

File tree

‎hack/e2e-chainsaw/clickhouse/chainsaw-test.yaml‎

Lines changed: 6 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -393,9 +393,10 @@ spec:
393393
# bucket's SeaweedFS and its Ingress controller sit.
394394
#
395395
# The precondition step deletes tenant-root's projected cozy-backups-creds,
396-
# which the mongodb and rabbitmq round-trips leave there after their
397-
# cozy-default BackupJobs. Nothing reads it between suites, and the projector
398-
# writes it again on the next BackupJob, which is the moment under test.
396+
# which every earlier BackupJob in tenant-root leaves there, the one of
397+
# clickhouse-2-backup-roundtrip included: the projector runs for each
398+
# BackupJob whatever its strategy. It writes the Secret again on the next
399+
# BackupJob, which is the moment under test.
399400
apiVersion: chainsaw.kyverno.io/v1alpha1
400401
kind: Test
401402
metadata:
@@ -547,7 +548,8 @@ spec:
547548
# the Ingress serving the Test certificate, plus one last 30s
548549
# attempt and its 10s pause
549550
# 480s BackupJob to Succeeded, including the driver's 2m hold
550-
# 480s RestoreJob to Succeeded, same hold
551+
# 480s RestoreJob to Succeeded; no hold by then, the stamp is
552+
# already older than the backup's own 2m wait
551553
# 90s the EXIT trap: Corefile replace plus a 60s Secret delete
552554
# 1330s (22m10s)
553555
# 25m leaves 2m50s for the SQL and kubectl calls between those waits, so

0 commit comments

Comments
 (0)