Forum

Hello, we have assisted in troubleshooting your dedicated server. The root cause of the connection refused error is not the service itself, but a node DNS resolution failure.

You have installed Tailscale on this server, which has taken over the system DNS (overwriting /etc/resolv.conf). However, there is currently no external nameserver configured on the Tailscale side, causing all queries to external domains (such as the image registry registry-1.docker.io) to return SERVFAIL.

As a result, the node cannot pull any container images. All Pods (including Zeabur's ingress and your service) are stuck in ContainerCreating, there is no service listening on the external port 443, and proxy connections are therefore refused.

Solution (choose one, both performed on your end):

1. Keep MagicDNS (Recommended): Go to the Tailscale admin console → DNS → Add a Global nameserver (e.g., 8.8.8.8) to allow external domains to resolve normally.
2. Restore system DNS directly: Run sudo tailscale set --accept-dns=false on the server. The downside is that this node will no longer be able to resolve other tailnet devices by name using MagicDNS.

Once configured, DNS will be restored, k3s will automatically re-pull the images, and the Pods and services will recover on their own within a few minutes. No redeployment is required. If they do not recover after a few minutes, please let us know.