Shorten kubectl jq command

Question

I am iterating over kubectl output that looks similar to this:

{
  "apiVersion": "v1",
  "items": [
    {
      "kind": "Pod",
      "metadata": {
        "name": "ubuntu1",
        "namespace": "development-namespace1",
      },
   ],
    {
      "kind": "Pod",
      "metadata": {
        "name": "ubuntu2",
        "namespace": "development-namespace2",
      },
   ],
}

I am trying to learn jq, but my command to get a value out of the first dictionary is really long:

kubectl get pods -o json -A | jq '.items | .[] | select(.metadata.name == "ubuntu1")  | .metadata.namespace'

How can I shorten this?

Answer 1

• This answer assumes:
  • Linux version: Debian 12 bookworm
  • jq version: 1.6
  • Kubernetes version: 1.28.3

---

Iterate over an array nested in a dict keyed by 'items'

Using this (abbreviated) json output dumped from kubectl get pods -o json -A...

{
  "apiVersion": "v1",
  "items": [
    {
      "kind": "Pod",
      "metadata": {
        "name": "ubuntu1",
        "namespace": "development-namespace1",
      },
   ],
    {
      "kind": "Pod",
      "metadata": {
        "name": "ubuntu2",
        "namespace": "development-namespace2",
      },
   ],
}

As @pmf said in a comment, you can select the value of a key nested in that array using this:

jq '.items[] | select(.metadata.name == "ubuntu1").metadata.namespace'

Explanation

• .items[] returns the array keyed by items
• select(.metadata.name == "ubuntu1"):
  • searches all the dicts
  • finds the dict matching the nested key: metadata.name and value: ubuntu1
• select(.metadata.name == "ubuntu1").metadata.namespace returns value of the .metadata.namespace in the selected dictionary

To reiterate, jq '.items[] | select(.metadata.name == "ubuntu1") returns:

{
  "metadata": {
    "name": "ubuntu1",
    "namespace": "development-namespace1",
  }
}

---

Bonus material, use jq for any shell command with columns

Parse 'ps' output

Assume you want to parse the output of the ps command...

Assume ps | <insert-sed-jq-pipe-here> returns the following...

$ ps | <insert-sed-jq-pipe-here>
    PID TTY          TIME CMD
  11899 pts/0    00:00:18 bash
2694259 pts/0    00:00:00 ps
2694260 pts/0    00:00:00 sed
2694261 pts/0    00:00:00 jq
$

You can convert any columnized text to json with:

jq -sR '[sub("\n$";"") | splits("\n") | sub("^ +";"") | [splits(" +")]]'

Parse 'ps' output as json

This example will dump the aforementioned ps as json...

$ ps | sed -ne '/PID/,$ p' | jq -sR '[sub("\n$";"") | splits("\n") | sub("^ +";"") | [splits(" +")]]'
[
  [
    "PID",
    "TTY",
    "TIME",
    "CMD"
  ],
  [
    "11899",
    "pts/0",
    "00:00:18",
    "bash"
  ],
  [
    "2694259",
    "pts/0",
    "00:00:00",
    "ps"
  ],
  [
    "2694260",
    "pts/0",
    "00:00:00",
    "sed"
  ],
  [
    "2694261",
    "pts/0",
    "00:00:00",
    "jq"
  ]
]
$

Parse 'top' output as json

You can even use this technique on top... but top looks a little tricky at first because:

• It's interactive by default
• You get extra text before the PID column header.... example:

$ top -bn 1 | head -n 8
top - 08:54:54 up 14 days,  1:13,  0 user,  load average: 0.11, 0.22, 0.30
Tasks:  96 total,   1 running,  95 sleeping,   0 stopped,   0 zombie
%Cpu(s):  0.0 us,  6.7 sy,  0.0 ni, 93.3 id,  0.0 wa,  0.0 hi,  0.0 si,  0.0 st
MiB Mem :   3767.1 total,    109.5 free,   2140.8 used,   1990.5 buff/cache
MiB Swap:   1024.0 total,    898.2 free,    125.8 used.   1626.4 avail Mem

    PID USER      PR  NI    VIRT    RES    SHR S  %CPU  %MEM     TIME+ COMMAND
2371290 root      20   0 1931856  48168  27024 S   6.7   1.2   0:30.64 flanneld
$

You can solve the interactive top problem with top -bn 1, which will run top once in non-interactive mode.

Finally, that same sed we used in the ps command will strip off the lines before the top PID... so the final top command as json is:

$ top -bn 1 | sed -ne '/PID/,$ p' | jq -sR '[sub("\n$";"") | splits("\n") | sub("^ +";"") | [splits(" +")]]'

... insert-top-json-output-here
$