Run a job on a local fleet

Let the machines on one network take their tasks straight from each other, with machines elsewhere joining over a peer link.

Open Clusters

Before you start

Open the cluster and find Network. Click Read the network. Machines on the same network are grouped as one island, with their local addresses. An island marked by address is a guess; after a run it says seen working directly.
Submit a job as usual (see Split a job across machines), or open one you submitted.
Click Run on the fleet. Clusters picks the hub: the online member with the most processors that is not a phone. Choose another under Hub if you prefer.
Clusters sends the hub and worker programs, starts them, and connects each worker: Local network for workers on the hub's network, Peer link for the rest, and Through Ocalt when nothing more direct answers.
Watch the run: each worker's tasks, its speed and how it is connected. Results come into the job as they arrive.
Click Stop the fleet to end it early. When every task is done the hub and the workers stop on their own.

A worker that goes offline does not lose work: the tasks it had taken go back to the queue after a short lease, and another worker runs them.

A fleet answers far faster than a task per command, so it suits jobs of many small tasks.
A phone can be a worker; keep the OcaltQL Client open on it while the run lasts.

Tell us which step stopped you and what you saw on screen. We answer from Support.