Monkube: Monitoring Kubernetes

This command-line utility lets you monitor different resources in different namespaces with ease

SUGGESTION: To make it easier to work with monkube, make the utility an executable (e.g., $ chmod u+x monkube) and add it to your $PATH.

Usage

Check the help option of Monkube to find the list of options

$ monkube --help
usage: Monkube [-h] [-r RESOURCES] [-n NAMESPACES] [-l NUM_LINES] [-s] [-t LABELS] [-nd {full,reduced}] [-ln NUM_NODES] [-tn TARGET_NODES] [-st STATUS] [-go] [-j JOBS]

Monitoring of kubernetes resources

options:
  -h, --help            show this help message and exit
  -r RESOURCES, --resources RESOURCES
                        Comma-separated list of resources to be monitored
  -n NAMESPACES, --namespaces NAMESPACES
                        Comma-separated list of namespaces to be monitored
  -l NUM_LINES, --num_lines NUM_LINES
                        Number of lines to print for each resource
  -s, --sort            Sort resources starting with the most recently used
  -t LABELS, --labels LABELS
                        Get subset of resources matching the specified comma-separated list of labels. Use '|' to specify multiple matches for a given key. EXAMPLE: "--labels key-1=value-1,key-2='value-2|value-3'"
  -nd {full,reduced}, --nodes {full,reduced}
                        Show allocation and utilization of nodes
  -ln NUM_NODES, --num_nodes NUM_NODES
                        Show data for the first 'num_nodes'
  -tn TARGET_NODES, --target_nodes TARGET_NODES
                        Comma-separated list of specific nodes to query
  -st STATUS, --status STATUS
                        Get pods matching 'status'
  -go, --gpu_only       Only show nodes with GPUs (use with -nd)
  -j JOBS, --jobs JOBS  Number of concurrent kubectl workers (default: 16)

Performance

monkube issues independent kubectl queries concurrently and parses their JSON with orjson:

• In --nodes full mode the cluster-wide pod listing is fetched one request per namespace in parallel rather than as a single --all-namespaces call. The apiserver/proxy is bound by per-stream throughput, so this is several times faster on large clusters (measured ~19s → ~6s on a 253-node / 9300-pod cluster).
• Per-(resource, namespace) table queries and the nodes/pods fetches run on a shared thread pool; use -j/--jobs to tune the worker count.
• JSON payloads are handed to orjson as raw bytes (no intermediate decode/rejoin). orjson is optional — if it is not installed the tool transparently falls back to the standard-library json module. Install it with pip install orjson.

Example

Here is an example of how to use the utility

$ monkube --resources pods,services --namespaces default,kube-system --num_lines 5 --nodes full --num_nodes 3
------------------------------------- PODS: ------------------------------------

NAMESPACE: 'default'
No resources found in default namespace.

NAMESPACE: 'kube-system'
NAME                                                 READY   STATUS    RESTARTS      AGE
event-exporter-gke-5b8bcb44f7-rfc78                  2/2     Running   0             24d
filestore-node-4lvts                                 3/3     Running   0             7h
filestore-node-4njxs                                 3/3     Running   0             15d
filestore-node-6gzt2                                 3/3     Running   0             20d
filestore-node-8xqzh                                 3/3     Running   0             26d

----------------------------------- SERVICES: ----------------------------------

NAMESPACE: 'default'
NAME         TYPE        CLUSTER-IP   EXTERNAL-IP   PORT(S)   AGE
kubernetes   ClusterIP   10.56.0.1    <none>        443/TCP   134d

NAMESPACE: 'kube-system'
NAME             TYPE        CLUSTER-IP    EXTERNAL-IP   PORT(S)         AGE
kube-dns         ClusterIP   10.56.0.10    <none>        53/UDP,53/TCP   134d
metrics-server   ClusterIP   10.56.10.80   <none>        443/TCP         134d

-------------------------------- CLUSTER NODES: --------------------------------

NAME: gke-vantroid-default-pool-06aeef3b-46q9
    cpu:   911.00 /   940.00 ( 96.91%)
    gpu:     0.00 /     0.00 (  0.00%)
    mem:     1.33 /     2.84 ( 46.90%)

NAME: gke-vantroid-default-pool-06aeef3b-6748
    cpu:   893.00 /   940.00 ( 95.00%)
    gpu:     0.00 /     0.00 (  0.00%)
    mem:     1.21 /     2.84 ( 42.56%)

NAME: gke-vantroid-default-pool-06aeef3b-bgbz
    cpu:   891.00 /   940.00 ( 94.79%)
    gpu:     0.00 /     0.00 (  0.00%)
    mem:     1.12 /     2.84 ( 39.56%)

--------------------------------------------------------------------------------

   ETCD time: 1.59 s
Parsing time: 0.01 s