Artem at Large

A quick one today: getting into your Kubernetes pods for debugging. Soon after having a cluster running you will run into odd situations where your application is mis-behaving and you need to inspect it’s environment. Thankfully kubectl exec lets you interactively run a command inside of a container and will stream the stdin/stdout back to your terminal. All you need to do is specify the exact pod and container to hook up to and kubectl will take care of the rest.

Our cluster had many replicas of the same pod. Most of the time it didn’t matter which pod in particular I wanted to inspect. After many rounds of running kubectl get pods | grep blah I wrote up a bash function to automate the process. It greps list of pods for the first argument specified and launches an interactive sh session in the default container. You can also specify the container name with the second argument.

Usage looks like:

$ kssh www
connecting to <pod id> ...

# SSH into first kubernetes pod with name $1
function kssh() {
  local pod="$(kubectl get pods
    | grep -E "^.*?$1.*?Running.*?$"
    | head -n 1
    | awk '{print $1}')"

  if [[ -z "${pod}" ]]; then
    echo "No such pods available"
  else
    local container=""
    if [[ ! -z "$2" ]]; then
      container="--container $2"
    fi
    echo "connecting to ${pod} $2..."
    kubectl exec -i -t "${pod}" ${container} /bin/sh
  fi
}

It may not seem like much but it has saved me many times when trying to debug a process, inspect network conditions in a cluster, or just inspect the environment variables that Kubernetes has injected into a pod.