Commit 2392326
committed
ci(openshift): pin router-default to control plane during E2E (cloudnative-pg#10734)
The drain_node E2E cordons one of three labelled worker nodes and drains
another, leaving the router-default Deployment's PDB unable to evict its
replicas onto the single surviving worker; the drain hangs and the test
fails intermittently on OpenShift.
Patch IngressController.spec.nodePlacement before the suite runs so
router-default lives on the control plane for the cluster's lifetime,
keeping all workers free for cordon and drain. The placement is
non-production, but the OCP cluster is destroyed at end-of-job.
Waits use observedGeneration plus rollout status, no fixed sleep.
Verified on an OCP 4.21 run of the equivalent change in the downstream
pg4k repo: the drain_node "3 pods on 2 nodes" spec (which had been
failing on the OCP nightly) is now green.
Signed-off-by: Marco Nenciarini <marco.nenciarini@enterprisedb.com>
(cherry picked from commit d0e5078)1 parent 999926e commit 2392326
1 file changed
Lines changed: 34 additions & 0 deletions
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
171 | 171 | | |
172 | 172 | | |
173 | 173 | | |
| 174 | + | |
| 175 | + | |
| 176 | + | |
| 177 | + | |
| 178 | + | |
| 179 | + | |
| 180 | + | |
| 181 | + | |
| 182 | + | |
| 183 | + | |
| 184 | + | |
| 185 | + | |
| 186 | + | |
| 187 | + | |
| 188 | + | |
| 189 | + | |
| 190 | + | |
| 191 | + | |
| 192 | + | |
| 193 | + | |
| 194 | + | |
| 195 | + | |
| 196 | + | |
| 197 | + | |
| 198 | + | |
| 199 | + | |
| 200 | + | |
| 201 | + | |
| 202 | + | |
| 203 | + | |
| 204 | + | |
| 205 | + | |
| 206 | + | |
| 207 | + | |
174 | 208 | | |
175 | 209 | | |
0 commit comments