Skip to content

Metrics

The platform continuously exposes operational signals about every processor it manages, in standard Prometheus format. These feed the platform's dashboards and alerts, so you can see at a glance whether your processors are healthy, how often they restart or move, and how the platform is scaling capacity underneath them.

You don't instrument any of this yourself — it comes from the platform's control plane. (Your processor also emits its own dory_* metrics from the SDK; see Observability.)

What the platform reports

The signals fall into a few plain-language groups:

Area What you can see
Processor fleet How many processors are running, pending, or shutting down, and how long they take to go from scheduled to running.
Health How many processors passed or failed their most recent health check, and whether the platform's own checks (database, cluster, config) are healthy.
Scheduling & placement Decisions the platform makes to provision, migrate, or decommission capacity, and how full each node is.
Edge & failover How many edge nodes are healthy, how many processors have been failed over from edge to managed capacity, and when edge-node failures occur.
Reconciliation How quickly the platform applies configuration changes, and any errors creating or moving processors.
SDK adoption Whether each managed processor is running the Dory SDK (so the platform knows it can coordinate graceful migration and state transfer).

These directly reflect the behaviors described elsewhere in this section — for example, the edge and failover signals correspond to Edge Failover, and the migration signals to State Migration.

How it's surfaced

All of the above is scraped by the platform's monitoring stack and rendered on its dashboards. Alerts are pre-configured for the conditions that matter most — the control plane being unreachable, the backing database being down, or processors taking too long to start or reconcile — so operators are notified before issues reach your processors.