Skip to main content

Description

List the model instances currently serving a model. The listing opens with a per-region table of DESIRED, RUNNING, HEALTHY and IDLE counts, printing ”?” for unknown values. A region where the model is scaled to zero is hidden from it unless --all is passed. Each instance below it reports its own STATUS: healthy (holding a fresh heartbeat), unhealthy (running, but not heartbeating yet), or terminated (served past sessions and now gone from its cluster, shown with --all). A ”?” means the server reported no status - an old server, or a cluster it could not reach - which is not the same as unhealthy. <model> is the model name (name or canonical org/name). With no argument, the name is inferred by walking up from the current directory to the nearest enclosing reactor.yaml’s model.name (stopping at the repository root). The default view is live-only. Pass --all to also include recent historical instances (those that served sessions but are no longer running); --since bounds that history window as a Go duration (plus a d day suffix), e.g. 30m, 24h, 7d - server default 24h, maximum 7d. Pass --json to emit the listing as JSON (useful for scripting). Examples:
Runtime errors (model not found, discovery unavailable) do NOT print the full usage screen - only argument / syntax errors do.

Options

Global options

See also