Lifecycle
Start, stop, and delete are three idempotent calls. A stop keeps everything, a delete keeps snapshots.
Start, stop, and delete are three idempotent calls. A stop keeps everything, a delete keeps snapshots.
Stop
Cuts a sync point, tears the pod down as soon as the cut is ready, and marks the workspace stopped. Seconds, and it never waits for a replica.
curl -sS -X POST https://dev.kloudlite.io/v1/workspaces/$WS/stop -H "Authorization: Bearer $KL_TOKEN"Start
Brings the pod back on the same tree. The most up-to-date node takes it, which is normally the node it was on.
curl -sS -X POST https://dev.kloudlite.io/v1/workspaces/$WS/start -H "Authorization: Bearer $KL_TOKEN"A start is refused with 409 when the workspace is running on a node that is down: it resumes when the node returns, or you clone it.
Delete
Removes the pod, the tree, and every sync point. Snapshots you pushed survive on a detached volume; with none, the volume goes too.
curl -sS -X DELETE https://dev.kloudlite.io/v1/workspaces/$WS -H "Authorization: Bearer $KL_TOKEN"What each keeps
| Tree | Home | Caches | Snapshots | |
|---|---|---|---|---|
| Stop | kept | kept | kept on the node | kept |
| Start on another node | newest sync point | kept | rebuilt | kept |
| Delete | gone | kept | gone | kept, on a detached volume |
Conditions
Three conditions on the document explain anything that is not plain ready:
| Field | Reason | Meaning |
|---|---|---|
replicated | Running, AwaitingReplica, Replicated | Whether another node holds the final sync point |
degraded | NodeDead | The node is down; a running workspace is interrupted |
decommissioning | NodeLeaving | The node is being retired; stop when convenient |