Hello, world (again)

This blog has been offline and online a few times since 2010, through WordPress, Ghost and Hugo. It’s back with a new look and a fresh start.

Everything I wrote before now is still here and marked as archived. Those posts are kept as they were written, so expect old versions of ESXi, OpenStack and friends. New posts start here.

Every hello world needs a hello world, so here’s one from a Kubernetes cluster.

Say hello

One line that greets you with every node and the kubelet version it runs:

kubectl get nodes -o json | jq -r '"hello, world from \(.items | length) nodes: " + ([.items[] | "\(.metadata.name) (\(.status.nodeInfo.kubeletVersion))"] | join(", "))'

text

hello, world from 3 nodes: cp-1 (v1.34.1), worker-1 (v1.34.1), worker-2 (v1.33.5)

A greeting that doubles as an upgrade check: worker-2 is a minor version behind.

And a few that earn their keep

Version skew at a glance. Same data, grouped, for clusters with more than a handful of nodes:

kubectl get nodes -o json | jq -r '[.items[].status.nodeInfo.kubeletVersion] | group_by(.) | map("\(.[0]) x\(length)") | join("  ")'

text

v1.33.5 x1  v1.34.1 x2

Who keeps restarting. Total restarts per pod across all its containers, worst first, zeros skipped:

kubectl get pods -A -o json | jq -r '.items[] | [([.status.containerStatuses[]?.restartCount] | add // 0), "\(.metadata.namespace)/\(.metadata.name)"] | select(.[0] > 0) | @tsv' | sort -rn | head

text

15	shop/api-7d9f
2	shop/web-55c

What’s actually unhappy. --field-selector=status.phase!=Running misses the classic case: a pod in CrashLoopBackOff still reports phase Running. This looks at container readiness instead, skips finished Job pods, and says why each pod is stuck:

kubectl get pods -A -o json | jq -r '.items[] | select(.status.phase != "Succeeded") | select((.status.containerStatuses // [{}]) | any(.ready != true)) | ["\(.metadata.namespace)/\(.metadata.name)", .status.phase, ([.status.containerStatuses[]? | select(.ready != true) | .state.waiting.reason // .state.terminated.reason // "NotReady"] | join(",") | if . == "" then "Unscheduled" else . end)] | @tsv' | column -t

text

shop/api-7d9f  Running  CrashLoopBackOff
shop/cache-0   Pending  Unscheduled
shop/img-x     Pending  ImagePullBackOff

All four need only kubectl and jq, and they’re read-only, so they’re safe to run against any cluster you can reach.

More soon.