Services replicas and scaling

A replica is one running copy of a service. A service runs from 1 to 50 replicas.

Change replicas, region and size

miosa service update shop API --replicas 4
miosa service scale  shop API --replicas 4

update changes the setting. scale changes replicas, region or limits and also queues a release that applies them. The running release is not touched until that release is active, so a scale change never interrupts traffic part way.

Setting	Meaning
`replicas`	1 to 50. A writable volume forces 1.
`region`	`us-west`, `us-east` or `us-mia`.
`limits`	`{vcpu, memory_mb}`. Above 4 vCPU or 8192 MB the release fails.
`restart_policy`	`always`, `on_failure` or `never`.
`healthcheck_path`	The path a web service must answer before a new release becomes active.

What a release does to replicas

A new release starts its replicas first. Once they pass the health check, the release becomes active and the earlier release drains and stops. overlap_s and terminate_grace_s add time before the earlier release stops.

Volumes

A volume is attached to a service, never added on its own. A writable volume allows one replica. Mount it read-only if you need more. A service can mount up to four volumes, and a path or a volume cannot repeat.

Check what takes effect

miosa service settings shop

Each setting is enforced, stored or unsupported. A stored setting is kept but not applied yet, so do not rely on it.

Where to go next

Run modes and types: every setting and which types it applies to.
Releases and rollback: how a scale change becomes a release.
Logs: what each replica prints.