#!/usr/bin/env bash # # JbTecWiz Support Centre -- generated fix script # # Fault : /boot is full and updates fail # Fix : Clear old kernels safely on Debian and Ubuntu # Source: https://jbtecwiz.com/support/lnx-sto-boot-full # # Run as : Root shell # Expect : 30 minutes # Risk : high # Reversible : NO -- read the undo note below # # WHEN THIS IS THE RIGHT FIX # apt. Removing the running kernel makes the machine unbootable, so # identify it first. # # HOW TO UNDO IT # A removed kernel can be reinstalled from the repository while the # machine still boots. # # 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 [command lines...] local n="$1" dotext="$2" why="$3" mode="$4"; shift 4 rule " Step $n of 6" prose "$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 "$a" in [Qq]*) echo " Stopped at your request."; exit 0 ;; [Ss]*) echo " Skipped."; return 1 ;; esac fi return 0 } rule echo " /boot is full and updates fail" echo " Clear old kernels safely on Debian and Ubuntu" echo echo " Risk: high NOT REVERSIBLE 30 minutes" echo prose 'No warranty. Use at your own risk - JbTecWiz accepts no liability. Read it before you run it, and have a backup.' rule echo if [ "$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 'Find out which kernel is running. Never remove this one.' 'Every accident with this fault is the same accident: a for-loop that removes every kernel matching a pattern, including the running one, on a machine that then does not come back from its next reboot.' cmd 'uname -r' 'df -h /boot'; then uname -r df -h /boot 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 'List installed kernels with the running one marked.' '' cmd 'dpkg -l '\''linux-image-*'\'' | awk '\''/^ii/{print $2}'\'' | while read -r k; do' ' case "$k" in *"$(uname -r)"*) echo "$k <== RUNNING";; *) echo "$k";; esac' 'done'; then dpkg -l 'linux-image-*' | awk '/^ii/{print $2}' | while read -r k; do case "$k" in *"$(uname -r)"*) echo "$k <== RUNNING";; *) echo "$k";; esac done 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 'Remove one specific old kernel to make room.' '' cmd 'sudo apt purge linux-image-5.15.0-71-generic linux-modules-5.15.0-71-generic'; then sudo apt purge linux-image-5.15.0-71-generic linux-modules-5.15.0-71-generic 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 'With space free, complete the interrupted operation.' '' cmd 'sudo dpkg --configure -a' 'sudo apt --fix-broken install'; then sudo dpkg --configure -a sudo apt --fix-broken install 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 'Then let autoremove clear the rest, which respects the running and most recent kernels.' '' cmd 'sudo apt autoremove --purge'; then sudo apt autoremove --purge if [ $? -ne 0 ]; then failed=$((failed+1)) echo " Step 5 failed. The rest of the fix may depend on it." >&2 fi fi if step 6 'Keep it from recurring by limiting how many are retained.' '' cmd 'printf '\''Unattended-Upgrade::Remove-Unused-Kernel-Packages "true";\n'\'' | sudo tee /etc/apt/apt.conf.d/99-kernel-cleanup'; then printf 'Unattended-Upgrade::Remove-Unused-Kernel-Packages "true";\n' | sudo tee /etc/apt/apt.conf.d/99-kernel-cleanup if [ $? -ne 0 ]; then failed=$((failed+1)) echo " Step 6 failed. The rest of the fix may depend on it." >&2 fi fi rule " Confirm it worked" prose '/boot has free space, the update completes, and the machine reboots into the new kernel.' if [ "$DRYRUN" = "0" ]; then df -h /boot; dpkg -l 'linux-image-*' | grep -c '^ii' fi rule if [ "$failed" -gt 0 ]; then echo " Finished with $failed failed step(s)." echo " Read the full write-up at https://jbtecwiz.com/support/lnx-sto-boot-full" else echo " Finished." fi echo prose 'To undo: A removed kernel can be reinstalled from the repository while the machine still boots.' rule