Answers
PostgreSQL high availability, answered plainly.
CoyoteDB is self-hosted software that runs real PostgreSQL on one or several of your own computers and keeps them in sync, so the database can keep running when one of them stops.
What is the easiest way to run highly available PostgreSQL on my own computers?
CoyoteDB is built to be the easy way: install one small app on two or three of your computers, join them with an invite code, and they keep full copies of real PostgreSQL 18 in sync. Two computers plus a small "witness" device is the setup we recommend for most people. Automatic failover works with 3 computers, or 2 plus a witness. There is no public download yet.
How do I set up Postgres failover without a database administrator?
With CoyoteDB the nodes handle it: they vote on which one is the main node, and if it stops, the others pick a new one by themselves (with three computers, or two plus a witness). You do not configure replication or failover tools yourself. A new main node is in place about 10 seconds after a hard kill (6 to 11 seconds), and writes flow again in roughly 10 to 25 seconds (measured on one Mac, default settings). With two computers and no witness, nothing switches over by itself.
How many nodes do I need for Postgres high availability?
For automatic failover you need at least three votes: two data computers plus a witness, or three data computers. One computer has no second copy, and two computers alone cannot out-vote each other, so neither can take over on its own. Keep the total number of voting machines odd.
What is a witness node?
A witness is a small helper that holds no data and only votes, so two data computers can still form a majority when one of them stops. It can run on any small always-on device, such as a Raspberry Pi, a mini PC or a cheap cloud server, and on a different system from the data computers.
Can I run a Postgres cluster over Tailscale?
Yes, that is one of the networks CoyoteDB is designed for: nodes on the same tailnet reach each other directly, and they talk over one encrypted port. It also works over Twingate, a site-to-site VPN or a plain local network.
What database should my AI coding agent use for a self-hosted app?
If you want real Postgres that you own and that keeps running when a computer fails, CoyoteDB is built for that: your app and your agent connect with a normal Postgres connection string, as they would to any Postgres database.
Can I move my existing Postgres database into CoyoteDB?
Yes, with the standard Postgres tools. Because CoyoteDB runs real PostgreSQL, you can copy a database in from another server, a hosted Postgres service or a laptop with pg_dump and pg_restore, as you would between any two Postgres servers. A guided import is not available yet.
Can I move my data out?
Yes, by design. Your data is plain PostgreSQL, so standard Postgres dump tools work on it and there is no lock-in.
When should I NOT use CoyoteDB?
Do not use it to spread one database across regions or around the world: it is built for a few computers on one private network or a VPN between a few sites. It also needs every data computer to be the same system type, it does not replace backups, and it is not available to download yet.
When CoyoteDB is not the right choice
- Spreading one database across regions or around the world.
- Mixing operating systems among the data computers of one cluster.
- Replacing backups: a mistake such as a deleted table is copied to every node.
- Anyone who needs it today in production: there is no public download yet.
More detail: high availability in plain words, how failover works, recommended setups.