*JSONPATH TIP* I don't remember if I shared this or not (the search history onl . . .

Kubernetes
#1

OE:
JSONPATH TIP

I don’t remember if I shared this or not (the search history only goes back a few days). So I’ll repost it

Say we want to list all OS images used by our nodes. Here are two ways we can tackle this challenge:

METHOD 1

we can use the trick mentioned here to flatten our JSON document and quickly search for paths: https://kubernetes.io/docs/reference/kubectl/cheatsheet/#viewing-finding-resources

so we would do the following:

$ sudo apt install jq

we’ll then use the trick mentioned here https://kubernetes.io/docs/reference/kubectl/cheatsheet/#viewing-finding-resources to flatten the document:

$ kubectl get nodes -o json | jq -c 'path(..)|[.[]|tostring]|join(".")' | grep -i osImage
"items.0.metadata.managedFields.4.fieldsV1.f:status.f:nodeInfo.f:osImage"
"items.0.status.nodeInfo.osImage"
"items.1.metadata.managedFields.4.fieldsV1.f:status.f:nodeInfo.f:osImage"
"items.1.status.nodeInfo.osImage"
"items.2.metadata.managedFields.4.fieldsV1.f:status.f:nodeInfo.f:osImage"
"items.2.status.nodeInfo.osImage"
"items.3.metadata.managedFields.4.fieldsV1.f:status.f:nodeInfo.f:osImage"
"items.3.status.nodeInfo.osImage"

we can further clean this up by removing the managedFields section:

$ kubectl get nodes -o json | jq -c 'path(..)|[.[]|tostring]|join(".")' | grep -i osImage | grep -v managedFields
"items.0.status.nodeInfo.osImage"
"items.1.status.nodeInfo.osImage"
"items.2.status.nodeInfo.osImage"
"items.3.status.nodeInfo.osImage"

so we can now easily write our query as follows:

$ kubectl get nodes -o jsonpath='{.items[*].status.nodeInfo.osImage}'

or if we wanted it in columnar format:

$ kubectl get nodes -o custom-columns=OS_IMAGE:.status.nodeInfo.osImage
OS_IMAGE
Ubuntu 18.04.5 LTS
Ubuntu 18.04.5 LTS
Ubuntu 18.04.5 LTS
Ubuntu 18.04.5 LTS

NOTE the above works in most cases but not all

METHOD 2 - VIM

here is a slightly longer method using Vim

$ kubectl get nodes -o json > /tmp/nodes.json
$ vim /tmp/nodes.json

we see a lot of output (over 2000 lines). Let us get to work:

:set ruler
:set foldmethod=syntax

this 2nd command folds our input into a single line

we then search for osImage. This opens only the paths that lead to osImages. Again, we ignore the managedFields section and look for the next hit. We see something like this (image1)

we split the screen and start writing the path:

status.nodeInfo.osImage

but we still need to build the rest of the path (before status). We can do this by looking at the bottom right and finding the column # (13)

:set colorcolumn=13

this adds a vertical ruler we can then use as we move up in our document to build the rest of our path (image 2). We move up and find that status is a child of items and we can now build our full query as we did in METHOD 1

json2.png
json2.png839×580 44.7 KB

json.png
json.png1004×408 47.7 KB

#2

BP:
Hi, Thanks for the JSONPATH tips. Appreciated. Are we allowed to perform sudo apt install jq in the exam?

#3

DN:
A very good stuff. Thanks a lot for sharing.

#4

Thanks for sharing. I find the part after path is hard to remember as I don’t use jq that much. You can get close enough with sed (which I use much more often than jq):

kubectl get SOMETHING ... -o json | jq -c paths | grep KEY_NAME | sed 's/","/./g'

eg

kubectl get pod nginx -o json | jq -c paths | grep startTime | sed 's/","/./g'

shows

["metadata.managedFields",1,"fieldsV1.f:status.f:startTime"]
["status.startTime"]

which is close enough to cut and paste in a jsonpath expression