Instances and the proxy
The proxy gives each new request to the next instance in turn. All instances share the work.
The proxy sends the first request to instance 1.
An instance is one container of a service. All the instances of a service do the same work.
The proxy shares the requests of a web service or a static site. The rule is round-robin: each instance gets a request in turn.
On the private network, the name of the service gives the addresses of all its instances. See Networking.
The limits are for each instance
Each instance has the memory and CPU limits of the service: 512 MiB and 1 CPU, unless you set other values. Thus 3 instances can use 3 times the limit.
ferry status shows the use of each instance. See Resource limits.
Manual scaling only
Ferry has no autoscaling. You choose the number of instances.
The state of a service is degraded while it has fewer instances than you asked for. An example is the moment after a crash. The state is live again when the reconciler replaced them.
See Logs & events.