Load balancing & failoverLoad balancing and failover across your servers.
Put several servers behind one hostname. nsin shares traffic between them, checks each one's health and takes a failing server out of rotation, so visitors never land on the dead one.
Origin pools
Add each server with its address, port and scheme, then choose how traffic is shared: round robin with weights, least load, or geographic, where each nsin location uses the server you assign it.
- Round robin with per-server weights
- Least-load balancing
- Geographic assignment per edge location
Health checks and automatic failover
nsin probes each server on a path you choose. Set the interval, the timeout, how many failures eject a server and how many successes bring it back. A server failing real requests can be ejected for a cool-off too.
- Active probes on / or a dedicated /healthz
- Failing servers leave rotation automatically
- Recovery without anyone being paged
Origin routes for split backends
Send matching paths to one server and leave the rest alone, such as /api/* to an application server. Override the Host header and TLS SNI when a backend serves several sites.
- Route by path to a specific server
- Host header override per rule
- Pools and routes can be combined
Frequently asked questions
What do visitors see when one server goes down?
With health checks on, nothing: traffic moves to the remaining servers in the pool.
Do I need a separate load balancer?
No. Origin pools run at the nsin edge. You add servers in a rule; there is no extra machine to run.
Should I use a pool or a route?
Several identical servers call for a pool. Different parts of a site on different servers call for a route. You can use both together.
What should a health check point at?
A path that exercises your application. A page that returns 200 while the database is down tells the check nothing.