kubeovn / kube-ovn / 26731493191

25%

master: 26%

Build Type

push

github

Committed by web-flow

Commit Message

fix(controller): skip in-pod cleanup when VPC NAT gateway CRD is deleted (#6784) (#6798)

Backport of #6784 to release-1.16.

The in-pod cleanup helpers (deleteEipInPod, delEipQoSInPod, deleteFipInPod,
deleteDnatInPod, deleteSnatInPod) used the NAT gateway pod as the termination
sentinel: a missing pod made delEipQoSInPod retry forever (EIP stuck) and made
the others silently skip cleanup even when the gateway was only being recreated.

Introduce the natGwDeleted helper that uses the VpcNatGateway CRD as the
sentinel: CRD gone -> return nil (nothing to clean up); CRD present but pod
absent -> return the error so the reconciler retries until the pod is ready
(logged at klog.V(4) to avoid noise during rolling updates).

Adapted from upstream commit f331bbb07: release-1.16 helpers take a single-arg
getNatGwPod(dp) (no namespace) and lack natGwNamespaceByName /
normalizeSnatInternalCIDR, so the changes were applied manually. Unit tests are
provided in a self-contained vpc_nat_gw_cleanup_test.go.

(cherry picked from commit f331bbb07)

Signed-off-by: Mengxin Liu <liumengxinfly@gmail.com>

Coverage Stats

63 of 82 new or added lines in 2 files covered. (76.83%)

17 existing lines in 1 file now uncovered.

14226 of 56669 relevant lines covered (25.1%)

0.29 hits per line