kuke refresh¶
Reconcile .status from the live runtime without touching .spec.
What it does¶
Walks every realm / space / stack / cell / container that Kukeon has metadata for, asks containerd and CNI what the actual state is, and updates the .status field of each resource on disk. It does not modify .spec or any runtime state.
Use it when:
- A container crashed and Kukeon's
statusis stillReady—refreshwill update it toFailed/Stopped. - You rebooted the host (or restarted containerd) and
kuke getstill shows pre-reboot state. - You intervened outside Kukeon (
ctr tasks kill,ip link delete) and want the metadata to catch up.
Flags¶
None beyond the global flags.
Output¶
A summary of what was inspected and what was updated:
$ sudo kuke refresh
Inspected 2 realms, 2 spaces, 2 stacks, 1 cells, 1 containers.
Updated 0 resources.
When it's a no-op¶
If Kukeon's metadata already matches the live state — i.e., you haven't rebooted, crashed, or intervened — refresh finds nothing to update. It's safe to run at any time.
refresh vs. get¶
getreads metadata and prints it. It never changes anything.refreshreads the live runtime and writes updated.statusto disk. The nextgetwill then reflect the new state.
So a typical "what's actually happening right now" workflow is: