osp_verification: purge bogus Machines and harden LB suite cleanup - #42
Conversation
ffb5ab7 to
cedfc1b
Compare
270a5a7 to
4cbb915
Compare
|
Follow-up from serval71 combo validation with openstack-test #321: Serial OTE was still marking stages UNSTABLE (rescued) when junit showed passed>0 and failed=0. Root cause: mid-loop treated run-test non-zero + empty per-chunk resolve as suite failure, even though the aggregated log resolved correctly (OTE often exits non-zero before a standalone chunk parses). Empty per-chunk resolve now falls back to ginkgo SUCCESS!/FAIL!/SKIP signals, and otherwise defers to the end-of-suite resolvable==expected check. Failed/unknown still fail immediately. |
|
|
||
| - name: Short wait after bogus Machine purge | ||
| ansible.builtin.pause: | ||
| seconds: 15 |
There was a problem hiding this comment.
Could we replace this fixed 15-second pause with a condition-based wait? The preceding deletions use --wait=false, so this task can continue while the Machines/MachineSets are still terminating or blocked by finalizers. openstack_test may then start with the leftover bogus-* resources still present, causing the exact contamination this cleanup is intended to prevent. Please poll until the matching resources are absent, with a bounded timeout.
There was a problem hiding this comment.
Agreed. Replaced the fixed 15s pause with a poll until no bogus-* Machines/MachineSets remain (30 retries x 10s, best-effort so a finalizer-stuck object does not hard-fail the stage). Amended tip.
There was a problem hiding this comment.
Follow-up: removed failed_when: false on the wait. If bogus-* Machines/MachineSets are still present after the bounded poll, the cleanup task fails and openstack_test does not start. Amended tip.
4cbb915 to
aeb6f67
Compare
Purge bogus-* leftovers before openstack_test, enable nightly LB guest cleanup including lb-sg-*, and fix serial OTE skip exits plus empty late sourceRanges junit. Serial OTE: ignore run-test non-zero only when resolver or ginkgo confirms skip/pass; empty per-chunk parse defers to end-of-suite resolvable count so pass/skip-heavy phases do not false-UNSTABLE. Bogus purge: poll until Machines/MachineSets are gone (bounded timeout) and fail if leftovers remain so openstack_test cannot start with contamination. Co-authored-by: Cursor <cursoragent@cursor.com>
aeb6f67 to
57e1c8c
Compare
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: ekuris-redhat The full list of commands accepted by this bot can be found here. The pull request process is described here DetailsNeeds approval from an approver in each of these files:
Approvers can indicate their approval by writing |
2 participants
Summary
bogus-*MachineSets/Machines so Machine Its are not poisoned by incomplete bz_2073398 cleanup (pairs with openstack-test fix; this is suite-level hygiene).lb-sg-*attachments so ProviderSpec is not tripped by LB leftovers when lb_tests runs first.Test plan
bogus-*in openshift-machine-api after openstack_test startsMade with Cursor