#!/usr/bin/env bash
#
# JbTecWiz Support Centre -- generated fix script
#
# Fault : "Connection refused" against "No route to host" against a timeout
# Fix   : Follow the packet until it stops
# Source: https://jbtecwiz.com/support/lnx-net-refused-vs-timeout
#
# Run as     : Both ends
# Expect     : 25 minutes
# Risk       : low
# Reversible : yes
#
# WHEN THIS IS THE RIGHT FIX
#   A timeout. Nothing answered, so the question is where it was dropped.
#
# HOW TO UNDO IT
#   Nothing was changed by these diagnostics.
#
# 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

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 "  "Connection refused" against "No route to host" against a timeout"
echo "  Follow the packet until it stops"
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 'Establish whether the packet arrives at all.' 'This settles it in one step. Packets arriving with no reply is a server or firewall problem on this machine; no packets at all is somewhere in between, and the two need completely different investigations.' cmd 'sudo tcpdump -ni any port 8080 -c 20'; then
  sudo tcpdump -ni any port 8080 -c 20
  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 'Trace the path if nothing arrives.' '' cmd 'mtr -rwzbc 20 10.0.0.20' 'traceroute -T -p 8080 10.0.0.20'; then
  mtr -rwzbc 20 10.0.0.20
  traceroute -T -p 8080 10.0.0.20
  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 local firewall for DROP rules.' '' cmd 'sudo nft list ruleset | head -60' 'sudo firewall-cmd --list-all 2>/dev/null'; then
  sudo nft list ruleset | head -60
  sudo firewall-cmd --list-all 2>/dev/null
  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 for asymmetric routing on a multi-homed machine -- a reply leaving by a different interface is silently dropped by reverse path filtering.' '' cmd 'ip route get 10.0.0.20' 'sysctl net.ipv4.conf.all.rp_filter'; then
  ip route get 10.0.0.20
  sysctl net.ipv4.conf.all.rp_filter
  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 'Check MTU if it connects and then stalls on larger transfers.' '' cmd 'ping -M do -s 1472 10.0.0.20 -c 3'; then
  ping -M do -s 1472 10.0.0.20 -c 3
  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 'tcpdump shows the handshake completing on both ends.'

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-net-refused-vs-timeout"
else
  echo "  Finished."
fi
echo
prose 'To undo: Nothing was changed by these diagnostics.'
rule