#!/usr/bin/env bash # # JbTecWiz Support Centre -- generated fix script # # Fault : /var full -- the journal or logs have consumed the disk # Fix : Reclaim the space and cap it permanently # Source: https://jbtecwiz.com/support/lnx-journal-disk-full # # Run as : Shell as root # Expect : 20 minutes # Risk : low # Reversible : yes # # WHEN THIS IS THE RIGHT FIX # /var is full. Do the capping step as well as the clearing step, or you # will be back here in a month. # # HOW TO UNDO IT # Vacuumed journal entries are gone permanently. The configuration # change is reversible by editing journald.conf back. # # 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 $n of 7" prose "$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 "$a" in [Qq]*) echo " Stopped at your request."; exit 0 ;; [Ss]*) echo " Skipped."; return 1 ;; esac fi return 0 } rule echo " /var full -- the journal or logs have consumed the disk" echo " Reclaim the space and cap it permanently" echo echo " Risk: low Reversible 20 minutes" echo prose 'No warranty. Use at your own risk - JbTecWiz accepts no liability. Read it before you run it, and have a backup.' rule echo if [ "$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 'Confirm where the space has gone.' '' cmd 'df -h /var' 'sudo du -xh /var --max-depth=2 2>/dev/null | sort -rh | head -15'; then df -h /var sudo du -xh /var --max-depth=2 2>/dev/null | sort -rh | head -15 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 'Check how much the journal is holding.' '' cmd 'journalctl --disk-usage'; then journalctl --disk-usage 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 'Trim it immediately to reclaim space.' '' cmd 'sudo journalctl --vacuum-size=200M'; then sudo journalctl --vacuum-size=200M 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 'Or trim by age instead.' '' cmd 'sudo journalctl --vacuum-time=7d'; then sudo journalctl --vacuum-time=7d 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 'Cap it permanently so this cannot recur -- set SystemMaxUse=500M in /etc/systemd/journald.conf.' 'With no cap, journald will use up to 10% of the filesystem. On a large /var that is a lot of logs and on a small one it is enough to fill it.' cmd 'sudo sed -i '\''s/^#\?SystemMaxUse=.*/SystemMaxUse=500M/'\'' /etc/systemd/journald.conf' 'sudo systemctl restart systemd-journald'; then sudo sed -i 's/^#\?SystemMaxUse=.*/SystemMaxUse=500M/' /etc/systemd/journald.conf sudo systemctl restart systemd-journald 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 'Check logrotate is actually running and not failing silently.' '' cmd 'systemctl status logrotate.timer' 'sudo logrotate -d /etc/logrotate.conf 2>&1 | tail -30'; then systemctl status logrotate.timer sudo logrotate -d /etc/logrotate.conf 2>&1 | tail -30 if [ $? -ne 0 ]; then failed=$((failed+1)) echo " Step 6 failed. The rest of the fix may depend on it." >&2 fi fi if step 7 'Find any single enormous log file that logrotate is not covering.' '' cmd 'sudo find /var/log -type f -size +100M -exec ls -lh {} \;'; then sudo find /var/log -type f -size +100M -exec ls -lh {} \; if [ $? -ne 0 ]; then failed=$((failed+1)) echo " Step 7 failed. The rest of the fix may depend on it." >&2 fi fi rule " Confirm it worked" prose '/var has free space and the journal is within its cap.' if [ "$DRYRUN" = "0" ]; then df -h /var && journalctl --disk-usage fi rule if [ "$failed" -gt 0 ]; then echo " Finished with $failed failed step(s)." echo " Read the full write-up at https://jbtecwiz.com/support/lnx-journal-disk-full" else echo " Finished." fi echo prose 'To undo: Vacuumed journal entries are gone permanently. The configuration change is reversible by editing journald.conf back.' rule