App commands

Durable apps are named workloads the daemon keeps a healthy instance of and re-creates from spec after a restart (see Apps). run is for throwaway work; app is for a server you want to stay up.

Command	Description
`create <name> --image <ref> [flags]`	create a durable app; prints its name
`update <name> [flags]`	replace the app's spec (same flags as create) and redeploy; zero-downtime for a proxy-fronted app (roll the new instance in, flip the route, drain the old); name immutable
`ls`	list apps (table: name, desired, phase, health, restarts, instance)
`get <name>`	full app JSON (desired state + observed status, incl. `instance_generation`)
`rm <name>`	delete the app and tear down its instance
`logs <name> [-f] [--source]`	the current instance's durable logs (`-f` reattaches across a redeploy)
`exec <name> [-i] [--cwd] [--timeout] [-e] -- <cmd>...`	run a command in the current instance
`shell <name> [--shell]`	interactive shell in the current instance
`sleep <name>`	snapshot the app and stop its VMM (free RAM+CPU), keeping its identity + ingress route; it wakes in place
`wake <name>`	wake a slept app (restore in place: same IP, clock stepped to now); idempotent on a running app

Create flags

--image (required), --pull, --restart always|on-failure|never, --health http:PORT[:PATH]|tcp:PORT, --health-cmd '<shell command>' (exec check, exit 0 = healthy), --port <guest port> (proxy target), -p/--publish (repeatable), -P/--publish-all (publish the image's EXPOSEd ports), -e/--env KEY=VALUE (repeatable, delivered to the entrypoint), --net-allow (repeatable), --net-allow-cidr (public IPv4 CIDR), --net-full-egress (any public host), --vcpus, --memory, --disk, --stopped, --idle-timeout <dur> (auto-sleep after this much idle time through the ingress proxy; 0/unset = never), --min-scale <n> (0 enables scale-to-zero; ≥1 keeps that many warm), --max-scale <n> (ceiling for horizontal autoscaling; > the floor enables it), --target-concurrency <n> (autoscaler's in-flight target per instance), --can-call <app> (repeatable; grant this app outbound access to <app>.internal — needs the daemon's --internal-networking), --internal-port PORT[/tcp|/http] (repeatable; expose a port to authorized peers at <app>.internal:PORT — tcp is a raw byte splice incl. TLS passthrough, http routes through the L7 proxy — needs the daemon's --internal-l4).

Operate flags

exec takes --cwd, --timeout <s>, -e/--env KEY=VALUE (repeatable), and -i/--interactive; shell takes --shell <path> (default /bin/sh); logs takes -f/--follow and --source service|exec|all.

crucible app create web --image nginx:alpine --port 80 -e LOG_LEVEL=info --restart always --health http:80:/
crucible app update web --image nginx:alpine --port 80 --memory 512   # rolls out zero-downtime
crucible app exec web -- /bin/sh -c 'nginx -t'
crucible app logs web -f                                              # reattaches if the app rolls
crucible app sleep web                                                # snapshot + free RAM; route kept
crucible app wake web                                                 # restore in place (same IP)
crucible app create api --image myapi --port 8080 --idle-timeout 5m --min-scale 0  # auto scale-to-zero
crucible app rm web