How to schedule something run once using anacron?

Question

Frequently within my scripts I schedule auxiliary tasks for later in the form:

at now + 8 hours <<< "rm -f workfile.$$ >> $worklog 2>&1 "

or

at 11pm + $(date +%M) minutes <<< "./postMarketJob.sh $orderID &> $orderID.log"

Trouble is, sometimes the laptops or PCs I use as servers are off when the scheduled time comes, so the at job never happens.

Same issue is true of crontab entries.

I'm trying to make sense of anacron's man page for Ubuntu.

Is there a tool to schedule a job such that it will run when requested, or at the earliest possibility whenever the server is back online?

It doesn't have to be with anacron, and if it's something only root can use, then I'll just wrap my invocations with sudo

CLARIFICATION

The tasks are not some fixed scripts (if at all) but very variable, and are conditional, meaning that depending on the unforeseen state of things within programs, they may or must not be scheduled in the first place.

It sounds like I might have to develop my own wrapper on top of atd/anacron that gives my programs a similar interface to at, but I was hoping something already exists. Been checking if fcron fits the bill and if available for Ubuntu in some repo.

Answer 1

For jobs that are executed once, at already does exactly what you require. At any time, atd will execute all scheduled jobs that are past their scheduled dates. In particular, at boot time, starting atd triggers the execution of any job whose date became due during the downtime.

For periodic jobs, you could combine a fixed-schedule crontab entry and an @reboot crontab entry. In the @reboot entry, only perform the task if it hasn't been done in the last 24 hours (or whatever interval you prefer).

42 5 * * * /usr/local/bin/myjob && touch /var/myjob/done
@reboot if [ -z "$(find /var/myjob.done -mtime 1)" ]; then /usr/local/bin/myjob && touch /var/myjob/done; fi

Answer 2

Anacron is great for what it is... a system to ensure jobs are executed every N number of days, after rebooting--not much control there. Furthermore, it's not a resident daemon and therefore can't replace crond or atd; it starts up at boot time (or when manually invoked), waits for the appropriate intervals before checking some conditions, and then once everything has been checked or run, it quits.

That said, it could be helpful in your situation, as a backup for atd. You could do something like:
echo '@monthly 0 custom1timejob script_or_cmd_here' >> /etc/anacrontab

Then schedule your at job, making sure to include something in the script that either lets anacron know the job was run (by doing date +%Y%m%d > /var/spool/anacron/custom1timejob), or even simpler: just remove the entry from anacrontab (sed -i /custom1timejob/d /etc/anacrontab).

Then, at worst, the job should be run the next time the system boots up.
(Make sure the RANDOM_DELAY and START_HOURS_RANGE variables in /etc/anacrontab [see man anacrontab], aren't gonna mess with your expectations though.)

Answer 3

This may be a situation where the batch option of at could help.

From my sandbox RHEL5 machine, I was able to schedule a batch job two minutes into the future, then stop atd. I let five minutes pass, and then started atd again. When atd started, the batch job kicked off immediately.

invocation: batch "NOW+2 minutes" or any acceptable at time input.

I did run into issues trying this on Fedora 16. Even though the man pages that say it supports time input, the command reports that it accepts no arguments.