#!/usr/bin/env bash
#
# JbTecWiz Support Centre -- generated fix script
#
# Fault : A container keeps restarting -- exit code 137, 139 or 1
# Fix   : Read what the application said before it exited
# Source: https://jbtecwiz.com/support/lnx-ctr-exit137
#
# Run as     : Root shell
# Expect     : 25 minutes
# Risk       : low
# Reversible : yes
#
# WHEN THIS IS THE RIGHT FIX
#   Exit code 1 or another application-level code.
#
# HOW TO UNDO IT
#   Diagnostics only.
#
# Walks the fix one step at a time and asks before each. Steps with no
# command are yours to do -- it prints those and waits. DRYRUN=1 prints
# without executing; UNATTENDED=1 does not ask.
#
# --------------------------------------------------------------------
# NO WARRANTY - USE AT YOUR OWN RISK
#
# This script is provided by JbTecWiz as-is and with no warranty of any
# kind, express or implied. You run it entirely at your own risk.
#
# JbTecWiz accepts no liability for any loss or damage arising from its
# use, including but not limited to data loss, downtime, or configuration
# changes that turn out to be wrong for your system.
#
# You are responsible for reading this script before running it, for
# satisfying yourself that it suits the machine in front of you, and for
# having a working backup first. Some steps cannot be undone.
# --------------------------------------------------------------------

set -uo pipefail
DRYRUN="${DRYRUN:-0}"
UNATTENDED="${UNATTENDED:-0}"
failed=0

if [ "$(id -u)" -ne 0 ]; then
  echo "  This fix is documented as needing root. Re-run with sudo." >&2
  exit 3
fi

rule() { printf "\n%s\n" "$(printf '-%.0s' $(seq 1 70))"; if [ $# -gt 0 ]; then echo "$1"; fi; }
prose() { echo "$1" | fold -s -w 74 | sed "s/^/  /"; }

# Returns 0 when the caller should run the command, 1 when it should not.
# A manual step always returns 1 -- there is nothing for the caller to run.
step() {  # step <n> <do> <why> <manual|cmd> [command lines...]
  local n="$1" dotext="$2" why="$3" mode="$4"; shift 4
  rule "  Step $n of 5"
  prose "$dotext"
  if [ -n "$why" ]; then echo; prose "$why"; fi
  if [ "$mode" = "manual" ]; then
    echo; echo "  -> Do this yourself, then press Enter to carry on."
    if [ "$UNATTENDED" = "0" ] && [ "$DRYRUN" = "0" ]; then read -r _; fi
    return 1
  fi
  echo; printf "    %s\n" "$@"; echo
  if [ "$DRYRUN" = "1" ]; then echo "  (dry run -- not executed)"; return 1; fi
  if [ "$UNATTENDED" = "0" ]; then
    read -r -p "  Run this step? [Y]es / [S]kip / [Q]uit " a
    case "$a" in
      [Qq]*) echo "  Stopped at your request."; exit 0 ;;
      [Ss]*) echo "  Skipped."; return 1 ;;
    esac
  fi
  return 0
}

rule
echo "  A container keeps restarting -- exit code 137, 139 or 1"
echo "  Read what the application said before it exited"
echo
echo "  Risk: low   Reversible   25 minutes"
echo
prose 'No warranty. Use at your own risk - JbTecWiz accepts no liability. Read it before you run it, and have a backup.'
rule
echo
if [ "$UNATTENDED" = "0" ] && [ "$DRYRUN" = "0" ]; then
  read -r -p "  Ready? [y/N] " go
  case "$go" in [Yy]*) ;; *) echo "  Nothing was changed."; exit 0;; esac
fi

if step 1 'Read the logs from the previous run, not the current one -- the container has already restarted.' '' cmd 'docker logs --tail 100 mycontainer' 'docker logs --tail 100 --since 10m mycontainer'; then
  docker logs --tail 100 mycontainer
  docker logs --tail 100 --since 10m mycontainer
  if [ $? -ne 0 ]; then
    failed=$((failed+1))
    echo "  Step 1 failed. The rest of the fix may depend on it." >&2
  fi
fi

if step 2 'In Kubernetes, the previous container'\''s logs are the ones that matter.' 'Without --previous you get the logs of the container that is currently starting, which has not failed yet. This is the single most useful flag for a CrashLoopBackOff and it is easy to miss.' cmd 'kubectl logs mypod --previous --tail=100' 'kubectl describe pod mypod | tail -20'; then
  kubectl logs mypod --previous --tail=100
  kubectl describe pod mypod | tail -20
  if [ $? -ne 0 ]; then
    failed=$((failed+1))
    echo "  Step 2 failed. The rest of the fix may depend on it." >&2
  fi
fi

if step 3 'Check the environment and configuration the container actually received -- a missing variable is the usual cause.' '' cmd 'docker inspect --format '\''{{range .Config.Env}}{{println .}}{{end}}'\'' mycontainer'; then
  docker inspect --format '{{range .Config.Env}}{{println .}}{{end}}' mycontainer
  if [ $? -ne 0 ]; then
    failed=$((failed+1))
    echo "  Step 3 failed. The rest of the fix may depend on it." >&2
  fi
fi

if step 4 'Check it can reach its dependencies. A database that is not up yet makes a container exit immediately and restart forever.' '' cmd 'docker run --rm --network container:mycontainer nicolaka/netshoot nc -zv db 5432'; then
  docker run --rm --network container:mycontainer nicolaka/netshoot nc -zv db 5432
  if [ $? -ne 0 ]; then
    failed=$((failed+1))
    echo "  Step 4 failed. The rest of the fix may depend on it." >&2
  fi
fi

if step 5 'Override the entrypoint to get a shell inside the failing image and test by hand.' '' cmd 'docker run --rm -it --entrypoint /bin/sh myimage'; then
  docker run --rm -it --entrypoint /bin/sh myimage
  if [ $? -ne 0 ]; then
    failed=$((failed+1))
    echo "  Step 5 failed. The rest of the fix may depend on it." >&2
  fi
fi

rule "  Confirm it worked"
prose 'The container starts cleanly and stays in the running state.'

rule
if [ "$failed" -gt 0 ]; then
  echo "  Finished with $failed failed step(s)."
  echo "  Read the full write-up at https://jbtecwiz.com/support/lnx-ctr-exit137"
else
  echo "  Finished."
fi
echo
prose 'To undo: Diagnostics only.'
rule