Reliable HTTP cron and delayed-job scheduling with automatic retries and full run logs — delay.run
You run crontab -e, add a line, save, and get this:
crontab: installing new crontab "/tmp/crontab.Yb3kQ2/crontab":4: bad minute errors in crontab file, can't install. Do you want to retry the same edit? (y/n) The good news: cron told you the line number (:4) and the field it didn't like. The bad news: the error messages are terse, and some of the worst crontab mistakes don't produce an error at all.
This post covers both.
Reading the error
The format is "<temp file>":<line>: <problem>. The line number counts every line in the file, including comments and blank lines. Answer y to reopen the editor with your changes intact, go to that line, and fix the field named in the message.
Cron lines have five time fields, then the command:
# ┌ minute (0-59) # │ ┌ hour (0-23) # │ │ ┌ day of month (1-31) # │ │ │ ┌ month (1-12 or jan-dec) # │ │ │ │ ┌ day of week (0-7 or sun-sat, 0 and 7 are Sunday) # │ │ │ │ │ 30 2 * * 1-5 /usr/local/bin/backup.sh The common messages
bad minute, bad hour, bad day-of-month, bad month, bad day-of-week: a value in that field is out of range or uses syntax cron doesn't know. Typical causes:
- Out of range:
60minutes,24hours,32days. Midnight is0, not24. - Unsupported characters:
L(last day),W(weekday),#(nth weekday),?. These come from Quartz and some cloud schedulers. Standard cron doesn't support them. - Mixed names and numbers in a range in ways your cron doesn't accept, or names in a step expression.
mon-friis fine; something likemon-5is risky. Stick to one style. - Stray spaces inside a list:
1, 15is two fields to cron. Write1,15.
bad command: cron found the time fields but no command after them, often because the line has only five fields. Also appears when a line was split across two lines by the editor or a copy-paste.
premature EOF or missing newline before EOF: the last line of the file doesn't end with a newline. Some editors don't add one. Go to the end of the last line and press Enter, or add a comment line at the bottom so the last real line is never last.
bad variable or errors on lines with =: environment lines like MAILTO=ops@example.com or PATH=/usr/local/bin:/usr/bin:/bin are allowed, but not every cron version handles spaces, quotes or variable expansion the same way. PATH=$PATH:/opt/bin doesn't expand $PATH in most crons; write the full value.
The mistakes cron accepts without complaint
These are worse, because crontab -e saves happily and the job then never runs, or runs at the wrong time.
Quartz or cloud syntax with six fields.
0 0 9 * * ? /usr/local/bin/report.sh This is "9 a.m. every day" in Quartz (seconds first). Standard cron reads 0 0 9 * * as "midnight on the 9th of every month", and the command as ? /usr/local/bin/report.sh, which fails at runtime. No install error.
Unescaped % signs. In a crontab, % in the command means "newline", and everything after the first % is sent to the command's standard input. This classic breaks silently:
0 3 * * * tar czf /backups/site-$(date +%F).tgz /var/www Escape it as \%, or move the command into a script.
Day-of-month and day-of-week together. 0 9 1 * 1 doesn't mean "the 1st, if it's a Monday". When both fields are restricted, cron runs when either matches: every Monday and the 1st of every month.
User field in the wrong place. Files in /etc/cron.d/ and /etc/crontab have an extra field, the user, between the schedule and the command:
30 2 * * * root /usr/local/bin/backup.sh Forget it there and cron tries to run as a user named /usr/local/bin/backup.sh. Add it in a personal crontab (crontab -e) and cron tries to run a command called root. Neither file is validated the way crontab -e validates, so you only find out from the logs.
Files in /etc/cron.d/ with the wrong name or permissions. Many cron implementations ignore files in cron.d whose names contain a dot (so backup.cron is silently skipped), and files that are group- or world-writable. Name them backup, owned by root, mode 0644.
The worst crontab bugs install cleanly and fail at 3 a.m. delay.run runs the job on an explicit schedule as an HTTP call, retries a failed run, and keeps a log of every attempt, so “did it run?” has an answer.
Check your crontab before it bites
A few habits that catch these early:
- Validate before installing. Keep the crontab in version control and install it with
crontab file.txt. The same validation runs, and a CI step can run it in a container on every change. - Preview the schedule. Paste the five fields into a cron expression explainer, or use a library like
croniter(Python) orcron-parser(Node) to print the next five run times:
from croniter import croniter from datetime import datetime it = croniter("0 9 1 * 1", datetime(2026, 11, 16)) print([it.get_next(datetime).strftime("%a %d %b %H:%M") for _ in range(5)]) - Log every run. Redirect output somewhere you'll read:
>> /var/log/jobs/backup.log 2>&1. Then check the cron log (grep CRON /var/log/syslogon Debian and Ubuntu,journalctl -u cronor-u crondelsewhere) for the line where cron started the command. - Put the logic in a script. One short line in the crontab calling a script avoids almost every
%, quoting andPATHsurprise.
Related reading
- Why '0 9 1 * 1' runs every Monday AND on the 1st: how cron combines day-of-month and day-of-week (and how to schedule 'the first Monday')
- Cron job not running? A step-by-step checklist (including “it works when I run it manually”)
- Cron expression for every Monday at 9am (and other weekly schedules)
- Cron in Docker FAQ: 11 questions from everyone whose job worked locally and does nothing in the container
- Free trials that convert to paid: a 12-point checklist for the scheduled jobs behind them (reminders, charges, downgrades)
Which crontab mistake has cost you the most time? For me it was a % in a date command that turned a nightly backup into a no-op for a month.
