Repository navigation
Commit 6f096b7
committed
fix(e2e): give the kafka teardown delete its own timeout
The verify-pvc-reclaim-on-delete step deletes the projected Kafka and
waits for it to return 404. There is no CR finalizer and no apps
controller in that path: apps.cozystack.io is served by an aggregated
apiserver that deletes the backing kafka-test HelmRelease synchronously
in the request handler, and the wait then blocks on finalizers.fluxcd.io,
which helm-controller holds until the release uninstall completes.
That uninstall is the heaviest teardown in the suite. Since #3938 it runs
a pre-delete hook Job first, then Strimzi tears down the broker and
ZooKeeper StatefulSets and their deleteClaim PVCs. On the normal path the
hook finishes in seconds and the Strimzi teardown dominates, landing on
the suite-default delete: 5m boundary (#4260). That default was sized for
finalizer-draining CRs elsewhere (VMDisk, FoundationDB, Tenant), so give
this one step finite headroom rather than widen the global.
8m does not stack the hook's 270s deadline on top of a full teardown: a
hook that burns that deadline means a wedged topic operator, where
redding is correct. So a genuinely stuck uninstall still fails loudly.
This is not the widen-instead-of-fix anti-pattern e2e-testing.md warns
against (that targets an unbounded hang, #3271) — the teardown is bounded
and #3938 already fixed the one hang that lived in this wait.
Refs #4260
Signed-off-by: Alexey Artamonov <[email protected]>1 parent b99db14 commit 6f096b7
1 file changed
Lines changed: 12 additions & 0 deletions
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
108 | 108 | | |
109 | 109 | | |
110 | 110 | | |
| 111 | + | |
| 112 | + | |
| 113 | + | |
| 114 | + | |
| 115 | + | |
| 116 | + | |
| 117 | + | |
| 118 | + | |
| 119 | + | |
| 120 | + | |
| 121 | + | |
111 | 122 | | |
| 123 | + | |
112 | 124 | | |
113 | 125 | | |
114 | 126 | | |
| |||
0 commit comments