jonaszi/debugging-shell-tool repository overview

Debugging Shell Tool is a small Alpine-based container image meant for temporary, in-cluster troubleshooting. It’s intentionally “idle” by default (sleeps forever), so you can deploy it into a Kubernetes cluster, kubectl exec into it, run your diagnostics, and then remove it again. This repo also includes Kustomize overlays and helper scripts to deploy either as a Deployment (single pod) or DaemonSet (one pod per node) and optionally target specific nodes / node pools.

Included tooling (in the image)

Shell: bash Network / DNS: curl, wget, bind-tools (dig/nslookup), mtr, traceroute, tcpdump, nmap, netcat (openbsd), socat IP / interface tooling: iproute2 (ip), iputils (ping), ethtool TLS / crypto: openssl, ca-certificates Process tooling: procps (ps/top/etc) JSON tooling: jq Misc: tzdata, ab (ApacheBench) What it’s for

DNS debugging inside the cluster (CoreDNS, stub domains, search paths) Connectivity checks between namespaces/services/pods Tracing routes and packet captures from the same network perspective as workloads Verifying node-level network characteristics (when deployed as a DaemonSet) Quick “is it the app or the network?” triage without baking tools into production images How it’s typically used in Kubernetes

Deploy temporarily into a target namespace Exec into the pod and run the tools you need Delete the controller again when finished (or let the provided shell script auto-delete on exit) Kustomize + scripts included in this repo

DaemonSet overlay (one pod per node): kustomize/overlays/daemonset Deployment overlay (single pod): kustomize/overlays/deployment Apply only: apply.sh Interactive shell with automatic cleanup on exit: shell.sh Delete/cleanup: delete.sh Notes / expectations

This is meant for short-lived debugging, not as a permanently running service. The image is intentionally broad in tooling; if you need a smaller footprint, you can build a reduced variant by removing packages from the Dockerfile. You can target specific nodes/node pools via node selectors (and for Deployment, optionally pin to a specific node).