Issue: Liveness probes fail on non-leader instances in HA mode
Issue: Liveness probes fail on non-leader instances in HA mode: a task in LegoFlow-SWE (Harbor dataset). Problem In HA deployments, the Argo Rollouts controller only starts the metrics server on the leader instance. Non-leader instances do not run the metrics server, causing their liveness probes…
The task
**Problem** In HA deployments, the Argo Rollouts controller only starts the metrics server on the leader instance. Non-leader instances do not run the metrics server, causing their liveness probes to fail because the metrics endpoint is unavailable. This leads to unnecessary restarts of non-leader pods.
Part of Lego-X/LegoFlow-SWE.