We have cluster with Istio and also Jenkins job to get "stable" pods, which uses this kubectl
query:
kubectl get po -o=jsonpath="{range .items[?(@.status.containerStatuses[-1].ready==true)]}{.spec.containers[0].image}{'\\n'}{end}"
registry/my-proj/admin:2.0.0.000123
registry/my-proj/foo:2.0.0.000123
registry/my-proj/bar:2.0.0.000123
This query fetches pods where last container (application) is ready, because we also have Istio sidecar containers. But here is tricky thing, it looks like array is built using alphabet, so if Istio container will be last - it fetches it as ready pod, because last container is ready.
I've tried to use go-template
also, but the best thing I've managed to do
kubectl get po -o go-template='{{range .items}}{{range .status.containerStatuses}}{{if eq .ready true }}{{end}}{{end}}{{.metadata.name}}{{println}}{{end}}
registry/my-proj/admin:2.0.0.000123
registry/my-proj/admin:2.0.0.000123
registry/my-proj/foo:2.0.0.000123
registry/my-proj/foo:2.0.0.000123
registry/my-proj/bar:2.0.0.000123
It fetches 2 times pods where 2 containers are ready and only 1 if 1 container is ready.
TL;DR;
I am looking for ultimate query which can fetch pods where all containers are ready, thanks
What about something like this?
kubectl get po -o go-template='{{range .items}}{{.metadata.name}}{{"\n"}}{{end}}{{range .items}}{{$pod := .}}{{range .status.containerStatuses}}{{if eq .ready false}}{{$pod.metadata.name}}{{"\n"}}{{end}}{{end}}{{end}}' | sort | uniq -u
What's happening here:
The trick is that -u
key excludes all duplicate entries, so all that is left is running pods.
{{ $pod := .}}
is used to save outer scope to print pod name inside inner loop. "Get all pods" is coming after "get not ready pods" to reduce risk of possible race condition, when some pods may have become ready by the time we did a "get all pods" query.
I do believe that something like this can easily be achieved with jsonpath too, but I don't think that you'll be able to do it with kubectl only, without the usage of sort
and uniq
.
If you are ok with grep
, you can use the following command:
kubectl get pod |grep -Po '^([^ ]+)(?=\s+((\d+)\/\3))'
Example:
kubectl get pod
NAME READY STATUS RESTARTS AGE
bar 2/2 Running 0 5m12s
foo 1/3 NotReady 6 6m9s
mypod 1/1 Running 2 (9m58s ago) 21h
kubectl get pod |grep -Po '^([^ ]+)(?=\s+((\d+)\/\3))'
bar
mypod
ps@controller:~$
Get all ready pods:
kubectl get po -o go-template='{{ $readyFlag := true}}{{range .items}}{{range .status.containerStatuses}}{{ if not .ready}}{{ $readyFlag = false}}{{end}}{{end}}{{if $readyFlag }}{{.metadata.name}}{{"\n"}}{{end}}{{end}}'
Get all not ready pods:
kubectl get po -o go-template='{{ $readyFlag := true}}{{range .items}}{{range .status.containerStatuses}}{{ if not .ready}}{{ $readyFlag = false}}{{end}}{{end}}{{if not $readyFlag }}{{.metadata.name}}{{"\n"}}{{end}}{{end}}'
Implementation details: 1. We define variable $readyFlag with initial value true 2. We loop over all containers statuses and if its field ".ready" is not true - update our flag value to false 3. In the end we just check if $readyFlag has value false - then at least one of checked containers had status not ready. If value true - then all containers was ready