Job states
States
Section titled “States”| State | Meaning | Terminal |
|---|---|---|
booked |
Accepted, with a node and a start time assigned | |
queued |
Waiting for capacity. No slot yet | |
running |
The container is up and spending tokens | |
finished |
Ran to completion, or the runtime expired | ✓ |
failed |
Stopped by an error - usually the image or the command | ✓ |
canceled |
Stopped by a user or an administrator | ✓ |
Service rows can also show transitional detail such as downloading image
while the node pulls it.
Transitions
Section titled “Transitions” ┌──────────► queued ──┐ (no capacity yet)submit ─────┤ ├──► running ──┬──► finished └──────────► booked ──┘ ├──► failed └──► canceled- booked → running when the start time arrives and the image is pulled.
- queued → booked when capacity frees up.
- running → finished when the command exits or the runtime expires.
- any non-terminal → canceled when someone cancels it.
Nothing leaves a terminal state. A second run is a new job, which is what Clone, Requeue and Restart produce.
What each state allows
Section titled “What each state allows”| Terminal | Logs | Extend | Cancel | Edit | Restart | Clone | |
|---|---|---|---|---|---|---|---|
booked |
✓ | ✓ | ✓ | ||||
queued |
✓ | ✓ | ✓ | ||||
running |
✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
finished |
✓ | ✓ | |||||
failed |
✓ | ✓ | ✓ | ✓ | |||
canceled |
✓ | ✓ |
How a job reports its services
Section titled “How a job reports its services”A job holding several services summarises rather than picking one: the row says how many are running. Expand it to see each service’s own state.
Common causes of failed
Section titled “Common causes of failed”| Cause | Tell |
|---|---|
| Image does not exist | manifest unknown in the logs |
| Private image, no registry selected | pull access denied |
| Command exits non-zero | The command’s own error |
| Out of memory | Killed shortly after start under load |