voodu stats

Synopsis

voodu stats [ref] [flags]

Single-shot (no streaming). Pipe through watch for a refresh loop.

One row per running replica — comparing siblings side by side is the fastest way to spot a leaky pod or an unbalanced load.

[ref] can be:

Empty: every running pod
A scope: myorg
scope/name: every replica of one resource
A kind: deployment / statefulset / job / cronjob
kind/scope/name: fully qualified

Common usage

# All pods in one scope
vd stats myorg

# One resource (one row per replica)
vd stats myorg/web

# All deployments across every scope
vd stats deployment

# Include leaks / pre-M0 legacy pods
vd stats --orphans

# Composable JSON
vd stats -o json

Output

$ vd stats clowk-vd
KIND        NAME                CPU        MEMORY
deployment  clowk-vd/docs.35a3  0.02/0.5   40.8Mi/100Mi
deployment  clowk-vd/docs.8f4c  0.001/0.5  40.6Mi/100Mi

Column	Meaning
`KIND`	`deployment` / `statefulset` / `job` / `cronjob`. `(orphan)` suffix means no matching manifest.
`NAME`	`scope/name.replica` — the operator's mental model.
`CPU`	used/limit, paired in the manifest's unit. Limit `0.5` → `0.45/0.5` (cores). Limit `500m` → `450m/500m` (millicores). No limit → `450m` (millicores fallback).
`MEMORY`	used/limit. Used comes from runtime bytes (rendered in `Mi`/`Gi`); limit echoes the manifest's verbatim per-pod string. `—` on either side = no data / no limit declared.

Flags

Flag	Description
`-k`, `--kind`	Filter by kind. Equivalent to the first segment of a positional ref.
`-s`, `--scope`	Filter by scope.
`-n`, `--name`	Filter by resource name.
`--orphans`	Include containers without a matching manifest (legacy / leaks). Hidden by default.
`-o`, `--output`	`text` (default), `json`, or `yaml`.

Positional ref and explicit flags are alternative spellings of the same filter — passing both is rejected.

Recipes

# Spot replicas approaching their memory limit
vd stats -o json | jq '.[] | select(.usage.memory_percent > 80)'

# Find a leaky replica — same resource, very different memory
vd stats myorg/web -o json | jq 'sort_by(.usage.memory_usage_bytes) | reverse'

# Find leaks (containers running but no manifest)
vd stats --orphans -o json | jq '.[] | select(.orphan)'

# Refresh loop (poor man's streaming)
watch -n 2 vd stats myorg

Notes

Stopped pods are omitted by design — no cgroup to sample. Use vd get pods to see them.
Per-pod rows are deliberate: averaging hides the asymmetry that diagnoses memory leaks, unbalanced load, or a wedged replica. If you need aggregated numbers, fold them yourself from JSON.
The JSON wire shape carries the full record: usage.cpu_percent, usage.memory_usage_bytes, usage.memory_percent, limits.memory_bytes, desired_replicas, etc.