Terminal UI for monitoring Kubernetes RayJobs and GPU resources. Think htop, but for your Ray-on-K8s cluster.
- RayJob-first view with drill-down into per-job pods
- GPU panel showing cluster-wide allocation, per-node sparkline, per-namespace breakdown
- Pending pods warning (cluster-wide, even when viewing a single namespace)
- Live-following logs with horizontal scroll (
[following]/[paused]indicator) - Kill RayJobs individually (
x) or by prefix (X) with confirmation - Ray Dashboard auto port-forward + browser open (
d) - kubectl exec into pods (
e) - Live search/filter (
/), sortable columns (s), namespace picker (n) - Auto-refresh every 2s
go install github.com/waleko/ktop/cmd/ktop@latestOr build from source:
git clone https://github.com/waleko/ktop.git
cd ktop
make installktop # current namespace
ktop -n my-namespace # specific namespace
ktop -A # all namespaces
| Key | Action |
|---|---|
tab |
Switch between RayJobs / Pods |
enter |
Drill into job pods / view pod logs |
l |
View logs (head pod on RayJobs tab) |
d |
Ray Dashboard (auto port-forward) |
e |
kubectl exec into pod |
x |
Kill selected RayJob |
X |
Kill all RayJobs matching prefix |
/ |
Search/filter |
s |
Cycle sort column |
n |
Namespace picker |
a |
Toggle all namespaces |
j/k |
Scroll up/down |
left/right |
Horizontal scroll (in logs) |
G |
Jump to bottom (in logs) |
q |
Quit |
kubectlconfigured with cluster access- KubeRay CRD installed (for RayJob features)
- Nodes with
nvidia.com/gpuresources (for GPU panel)
MIT