Forum

Hi,

The crash happens because the container runs headscale without the serve subcommand — it prints the help text and exits immediately, causing the restart loop.

To fix this, go to your Headscale service Settings and set the Custom Start Command to:

headscale serve

Then restart the service. It should start normally after that.

Since this template was created by another user, we're unable to fix it on the template side. If you plan to use Headscale long-term, you can create your own template with the correct start command configured, so future deployments work out of the box.

If you have any other questions, please create a new post. Thanks!

Hi, thanks a lot. But it still seems not work.

Hi,

Thanks for the screenshots. I checked the service logs — the start command fix did take effect, so the original issue is resolved. However, the service is now crashing with a different error:

No config file found, using defaults
FTL Error initializing: headscale now requires a new noise.private_key_path field in the config file for the Tailscale v2 protocol
Fatal config error: dns.nameservers.global must be set when dns.override_local_dns is true

Headscale v0.28.0 requires a configuration file with specific fields (noise.private_key_path, dns.nameservers.global, etc.) to start. You'll need to create a config file and mount it into the container for the service to run. You can refer to the official Headscale configuration documentation for details.

Thanks a lot. let me try it