Hacker News Reader

I wonder if this will fix the issue of appending -n to new ephemeral servers that join the network. For example, if you have a service wiki-1 and that container/instance gets restarted, it then appears on your tailnet as wiki-1 making users unable to access it at wiki/

Their official solution is to run a logout command before shutting down but that's not always possible.

View on HN

This is where mounting a state volume helps. If you mount a state volume, you don't need to make the containers ephemeral unless you expect them to move between hosts frequently. If that is the case, I'd love to hear more about your use case so that I can suggest a better alternative.

That is helpful to know for Docker containers, thank you.

The use-case where we find the renaming most frustrating is typically when we start a cloud instance with a Tailscale setup script in the cloud init (via Terraform). If we, say, change a parameter that requires Terraform to restart that instance, then the freshly-started instance will be given a `-1` name by Tailscale and the old instance will be offline.

I wish there would simply be a --force-hostname option or something of that nature that tells Tailscale "if a host is authenticating with this name, give it that name, any older machines using that name should be kicked off"

Reply on news.ycombinator.com