#!/usr/bin/env bash # # JbTecWiz Support Centre -- generated fix script # # Fault : A cron job works by hand but not on schedule # Fix : Give the job an environment and somewhere to complain # Source: https://jbtecwiz.com/support/lnx-cron-not-running # # Run as : Shell # Expect : 25 minutes # Risk : low # Reversible : yes # # WHEN THIS IS THE RIGHT FIX # Any cron job that works interactively and not on schedule. # # HOW TO UNDO IT # crontab -e to restore the previous entry; keep a copy with crontab -l # > ~/crontab.bak first. # # 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 [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 "$a" in [Qq]*) echo " Stopped at your request."; exit 0 ;; [Ss]*) echo " Skipped."; return 1 ;; esac fi return 0 } rule echo " A cron job works by hand but not on schedule" echo " Give the job an environment and somewhere to complain" echo echo " Risk: low Reversible 25 minutes" echo prose 'No warranty. Use at your own risk - JbTecWiz accepts no liability. Read it before you run it, and have a backup.' rule echo if [ "$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 'Prove cron is even running the job.' '' cmd 'sudo systemctl status cron 2>/dev/null || sudo systemctl status crond' 'sudo journalctl -u cron -u crond --since '\''2 hours ago'\'' --no-pager | tail -30'; then sudo systemctl status cron 2>/dev/null || sudo systemctl status crond sudo journalctl -u cron -u crond --since '2 hours ago' --no-pager | tail -30 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 'Capture the output -- by far the most useful single change. Without this the job fails into nothing.' '' cmd 'crontab -l'; then crontab -l if [ $? -ne 0 ]; then failed=$((failed+1)) echo " Step 2 failed. The rest of the fix may depend on it." >&2 fi fi step 3 'Redirect both streams to a log file in the crontab entry, e.g. `*/5 * * * * /opt/scripts/job.sh >> /var/log/job.log 2>&1`.' 'cron mails output to the local user by default, and on a machine with no MTA that mail goes nowhere at all. Redirecting is what makes the failure visible.' manual || true if step 4 'See exactly what environment the job gets by scheduling a one-off dump of it.' '' cmd '( crontab -l; echo '\''* * * * * env > /tmp/cron-env.txt 2>&1'\'' ) | crontab -' 'sleep 70 && cat /tmp/cron-env.txt'; then ( crontab -l; echo '* * * * * env > /tmp/cron-env.txt 2>&1' ) | crontab - sleep 70 && cat /tmp/cron-env.txt 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 'Use absolute paths for every binary, or set PATH at the top of the crontab.' '' cmd '( echo '\''PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin'\''; crontab -l | grep -v '\''^PATH='\'' ) | crontab -'; then ( echo 'PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin'; crontab -l | grep -v '^PATH=' ) | crontab - 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 'Escape any percent signs -- cron treats an unescaped % as a newline, which silently truncates the command.' 'This one catches almost everybody with a date-stamped filename. The command runs, just not the whole of it.' cmd 'echo '\''0 2 * * * /usr/bin/mysqldump db > /backup/db-$(date +\%F).sql'\'''; then echo '0 2 * * * /usr/bin/mysqldump db > /backup/db-$(date +\%F).sql' 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 'For anything non-trivial, prefer a systemd timer -- it logs to the journal, has proper dependencies, and can be tested on demand.' '' cmd 'systemctl list-timers --all | head'; then systemctl list-timers --all | head 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 'The job runs on schedule and the log file shows its output.' if [ "$DRYRUN" = "0" ]; then tail -20 /var/log/job.log fi rule if [ "$failed" -gt 0 ]; then echo " Finished with $failed failed step(s)." echo " Read the full write-up at https://jbtecwiz.com/support/lnx-cron-not-running" else echo " Finished." fi echo prose 'To undo: crontab -e to restore the previous entry; keep a copy with crontab -l > ~/crontab.bak first.' rule