Skip to content

Runtimes

Open it: Runtimes · under Runtimes & runs in the console menu.

Runtimes shows what is running and where: each FlowR server (a host), the runtimes that take work, and the deployments on each, with their health and load.

Runtimes: each FlowR with its node's memory, and the runtimes with what is deployed

Items is a tree: hosts, then runtimes, then what is deployed on each.

Each copy of FlowR reports itself every 20 seconds; Items lists every one seen in the last minute.

On a hostWhat it means
Node memoryMemory used on the machine it runs on, with its CPUs and load
This FlowR, CPU and memoryWhat this copy of FlowR uses, out of what it may use (its pod’s limits)
DiskIts data disk
Runtimes on itThe runtimes, with how many deployments each has; no worker on its pool marks one that nothing is taking work for
ControllerWhether the EE controller answers: when it does, each deployed runtime gets its own pods

Work pools lists the queues runs wait in until a worker claims them, with whether a worker is live and how many runs are queued.

A runtime is a place deployed work runs: by default the shared runtime rt-default, or one made for a project, an organisation or a single agent. A runtime carries labels (such as class=cpu, tier=prod, a3t.io/org=acme) that canvases select it by; see Canvas.

ControlWhat it doesWho
PauseIt takes no new work; callers get its paused responseA project’s writers for its own runtime; an organisation’s admins for theirs; SuperAdmin for shared ones
ResumeIt takes work againAs above
HibernateStops its pods until work arrivesAs above
TerminateRemoves it. First shows what is deployed there and where each moves; confirm with its nameAs above

New runtime registers one: a name, a project, its labels.

Each deployment on a runtime shows whether it is serving, its runs and failures. Its controls: Pause (with a reason, shown to callers), Resume, Restart, Undeploy.