The eight states
deleting is not a step in the flow — it is a lock. While it is held, no other
operation touches that agent.
The four actions
Deploy
On a
draft or stopped agent. Provisions a container and starts it. The
volume from a previous life is reattached, so the agent picks up where it
left off.Stop
On a
running agent. Removes the container and keeps the volume. Billing for
the seat does not stop — the seat is what you bought.Start
On a
stopped agent that still has its container definition. Faster than a
full deploy.Delete
Destroys the container and the volume. See
Deleting an agent — it asks you to type the name.
Waiting for capacity
When the platform has no room for another container, a deploy does not fail — it queues. The agent’s page shows a capacity wait panel with your position, how long you have waited, how long the seat is held for, and a way to stop waiting. When room appears, agents that are waiting start on their own; you get a notification rather than having to come back and check.When an agent breaks
The failure notice sits above everything on the agent’s page and is classified from what the container actually reported:
Configuration failures are the ones you can fix yourself — see
The agent’s config file. The others carry a contact-support
action with the agent id and the classification already filled in.