Border0 connector high availability

The Border0 provisioning system is designed with redundancy and high-availability in mind.

Connectors and sockets have a many-to-many relationship. A single connector can provide access to multiple sockets. A single socket can be served by multiple connectors. We recommend configuring at least two connectors for every production socket. Multiple connectors provide high availability by allowing the socket to remain accessible if a connector or its host becomes unavailable.

You can deploy connectors in different availability zones, cloud compute regions, or even cloud compute provider infrastructures. For example, you could have one cloud region that has three connectors that service ten sockets. Use your availability requirements to decide where and how many connectors you should deploy.

When happens when a connector fails?

All client devices are peered with all relevant connectors at all times, even if no traffic is being sent to them. When a connector is shut down or when the Tailscale control plane detects that it is offline, all client devices will be signalled regarding the event. If your sockets are served by another connector that is available, the clients will automatically use that connector.

What happens to existing connections when a connector goes down?

For sockets such as SSH, database, HTTP/HTTPS, or TCP servers, your current session ends, but a reconnect (or automatic reconnect, depending on the client) transparently uses the next available connector.