kubeovn / kube-ovn / 24754856824

26%

push

github

fix(speaker): require NODE_NAME and inject POD_IPS in v2 chart (#6669)

* fix(speaker): require NODE_NAME and inject POD_IPS in v2 chart

Follow-up to #6664. Two related hardening changes:

1. Add POD_IPS env to the kube-ovn-v2 speaker DaemonSet so dual-stack
   nodes expose both v4 and v6 pod addresses, matching yamls/speaker.yaml
   and the fallback logic in pkg/speaker/config.go.

2. Make an empty --node-name fail the speaker at startup instead of
   silently skipping every pod under the BGP "local" policy. The check
   is added to validateRequiredFlags so the error is aggregated with
   other missing flags.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Signed-off-by: Mengxin Liu <liumengxinfly@gmail.com>

* fix(speaker): skip --node-name requirement in NAT GW mode

Addresses Copilot review feedback on #6669.

The BGP speaker runs in two modes:
  - regular: syncSubnetRoutes reads config.NodeName to match the BGP
    "local" policy against pod.Spec.NodeName.
  - --nat-gw-mode: syncEIPRoutes only, never reads NodeName.

util.GenNatGwBgpSpeakerContainer injects GATEWAY_NAME / POD_IP / POD_IPS
but no NODE_NAME, which would make the previous unconditional check
crash every VPC NAT GW speaker pod at startup. Gate the requirement on
!NatGwMode and add a regression test case covering NAT GW mode with
empty NodeName.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Signed-off-by: Mengxin Liu <liumengxinfly@gmail.com>

---------

Signed-off-by: Mengxin Liu <liumengxinfly@gmail.com>
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

3 of 3 new or added lines in 1 file covered. (100.0%)

14015 of 56572 relevant lines covered (24.77%)

0.29 hits per line