#!/usr/bin/env bash
#
# JbTecWiz Support Centre -- generated fix script
#
# Fault : LVM thin pool full -- volumes go read-only or the machine hangs
# Fix   : Reclaim space when there is none left to add
# Source: https://jbtecwiz.com/support/lnx-sto-lvm-thin
#
# Run as     : Root shell
# Expect     : 40 minutes
# Risk       : high
# Reversible : NO -- read the undo note below
#
# WHEN THIS IS THE RIGHT FIX
#   The volume group is also full.
#
# HOW TO UNDO IT
#   Removed snapshots cannot be recovered.
#
# 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 "  LVM thin pool full -- volumes go read-only or the machine hangs"
echo "  Reclaim space when there is none left to add"
echo
echo "  Risk: high   NOT REVERSIBLE   40 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 'List the snapshots -- these are almost always where the space has gone.' 'A snapshot taken for a backup and never removed keeps every block the origin has changed since. A month of those will consume a pool on its own, and removing them is instant relief.' cmd 'sudo lvs -a -o lv_name,lv_attr,origin,data_percent,lv_size | grep -E '\''^\s+\S+\s+s'\'''; then
  sudo lvs -a -o lv_name,lv_attr,origin,data_percent,lv_size | grep -E '^\s+\S+\s+s'
  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 'Remove the snapshots you no longer need, oldest first.' '' cmd 'sudo lvremove vg0/snap-2026-05-01'; then
  sudo lvremove vg0/snap-2026-05-01
  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 'Enable discard so deletions inside the guest filesystem return blocks to the pool.' 'Without discard passdown, deleting a file inside a thin volume frees nothing in the pool. This is why a pool keeps filling on a filesystem that has plenty of free space.' cmd 'sudo lvchange --discards passdown vg0/thinpool' 'sudo fstrim -av'; then
  sudo lvchange --discards passdown vg0/thinpool
  sudo fstrim -av
  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 'Add a physical volume to the group if there is a spare disk.' '' cmd 'sudo pvcreate /dev/sdc' 'sudo vgextend vg0 /dev/sdc' 'sudo lvextend -l +100%FREE vg0/thinpool'; then
  sudo pvcreate /dev/sdc
  sudo vgextend vg0 /dev/sdc
  sudo lvextend -l +100%FREE vg0/thinpool
  if [ $? -ne 0 ]; then
    failed=$((failed+1))
    echo "  Step 4 failed. The rest of the fix may depend on it." >&2
  fi
fi

step 5 'Set up monitoring on Data% -- this fault has no gentle warning stage.' '' manual || true

rule "  Confirm it worked"
prose 'Data% has fallen and fstrim reports blocks returned.'
if [ "$DRYRUN" = "0" ]; then
  sudo lvs -a; sudo fstrim -av
fi

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-sto-lvm-thin"
else
  echo "  Finished."
fi
echo
prose 'To undo: Removed snapshots cannot be recovered.'
rule