Forum

Hello,

We have processed this for you from the backend. All 9/9 system components are now back to normal operation, and your stock-zeabur service has been restarted.

Why "Reinstall" didn't work

It wasn't that the button was unresponsive, but that the issues on your server were deeper than what the reinstall process could handle:

1. The old ingress-controller pod was stuck in the Terminating state for several hours due to a stale finalizer that wouldn't clear. Every time you clicked "Reinstall Zeabur Service," the Zeabur backend process would wait for the old pod to terminate before creating a new one. Since the old one never terminated, it would time out after 10 minutes, fail, and the health warning would reappear.
2. The k3s control plane itself was timing out (the API server kept returning http: Handler timeout). This meant that even the DaemonSet/Deployment controllers couldn't receive responses from the API server, so new pods couldn't be created.
3. Due to this, the vector-aggregator deployment accumulated 11 duplicate ReplicaSets. Each reinstall created a new RS, but the pods remained in a Pending state, resulting in a pile of leftovers.

What we did

• Force-deleted the stuck Terminating ingress-controller pod.
• Cleared all stuck vector-aggregator Pending pods and ReplicaSet leftovers.
• Cleared 16 service pods that had been Evicted due to memory pressure.
• Restarted the k3s service to reset the control plane.

Current Status

• All 9 system components are Running (the 6 anomalies previously displayed have been resolved).
• Your stock-zeabur service pod has been rebuilt and is 1/1 Ready.
• Memory: 891MB available (2GB limit), status is healthy.

Recommendations

If the system health warning appears again in the future, please do not repeatedly click "Reinstall Zeabur Service." In cases where underlying finalizers are stuck or the control plane is timing out, the reinstall button cannot fix the issue and will only leave behind more residual pods. Please reply to this thread or open a new ticket directly, and we will intervene from the backend, which will be much faster.