Deployment statuses

Every value of `DeploymentStatus`, what it means on Skiffly, what you can do in that state and how the dashboard, CLI and MCP show it.

The DeploymentStatus enum is Railway-compatible; the CLI and MCP print the lowercase form.

Status	Meaning	Actions
`QUEUED`	Accepted, waiting for a node to pick it up	cancel
`BUILDING`	Image build running, or waiting for one of the two build slots on the node (`waiting for a build slot`)	watch build logs
`DEPLOYING`	Image ready, containers starting, health check not passed yet; the previous deployment still serves traffic	watch runtime and system logs
`SUCCESS`	Live. For a cron service: the schedule is registered	redeploy, restart, roll back to it later
`FAILED`	Build error, health check not passed in time (see limits), image pull error, quota exceeded, or node lost for 20 minutes; the previous deployment keeps running	read `message` and logs, fix, deploy
`CRASHED`	The container exited repeatedly after going live (3 restarts in a row), or the restart policy stopped it after `max retries` crashes (scaled to zero, `stopped after N restarts`); `message` contains the last log lines	read logs, fix, deploy; restart; roll back
`SLEEPING`	Scaled to zero by App Sleeping; the next HTTP request wakes it	send a request, or turn sleeping off
`REMOVED`	Replaced by a newer deployment, cancelled, removed by hand, or the service was deleted	roll back (its saved image is redeployed, no rebuild, while `canRollback` is true — up to 14 days after removal)

Reserved and never returned today: INITIALIZING, NEEDS_APPROVAL, REMOVING, SKIPPED, WAITING.

Messages#

Deployment.message explains waits and failures with the underlying event:

Message	Cause
`waiting for a build slot`	two builds already running on the node
`exceeded quota …`	the environment's CPU/memory quota is full; lower limits or replicas of other services
`ImagePullBackOff …`	the image name or tag is wrong, or the registry is private
`OOMKilled`	the container used more than its memory limit
`CrashLoopBackOff: …`	the process keeps exiting; the last log lines follow
`build timed out / node lost`	no status from the node for 20 minutes
`cancelled by <user>`	cancelled while queued

Dashboard: service card on the canvas, Deployments tab, project Logs.
CLI: skiffly status, skiffly deployment list, skiffly up (streams until SUCCESS or a failure).
MCP: get-status, list-deployments, get-logs.
GraphQL: deployment(id) { status message }, deployments(input: { status: { in: [FAILED, CRASHED] } }).