The issue you're experiencing is related to the kube-ovn-controller not being ready when ENABLE_METRICS is set to false. This is because the health check for the controller relies on port 10660, which is not opened when metrics are disabled. The readiness probe is configured to use this port for health checks, and if the port is not open, the probe will fail, causing the controller to be marked as not ready.

In contrast, the CNI server logic opens port 10665 even when ENABLE_METRICS is set to false, which is why it does not face the same issue.

To resolve this, you would need to modify the controller's logic to ensure that the health check port (10660) is opened regardless of the ENABLE_METRICS setting, similar to how the CNI server handles its port. This would involve changing the conditional logic in the controller's code to allow the health check server to run independently of the metrics server.

To continue talking to Dosu, mention @dosu.

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Other}

创建server的时候会校验ExtraHandlers是否覆盖了默认的路径 “/metrics”

start 会默认监听 “/metrics”