launchd ignores RunAtLoad=false

Thanks, twtwtw for your reply. The intention of the KeepAlive key is to keep the job running until the backup is done successfully (which also does not work but I guess that would be another disussion thread). I have tried to take it out but that did not help. The script acutally really does what it should, it also runs at the specified times. It just runs ALSO at the time when the launchd plist is loaded, i.e. daily when I start up my Mac, which is weired and unnesseccary.

@VikingOSX What do you mean with 're-organize'? It is in exactly the order you propose ...

Is there a difference between the keys 'Program' and 'ProgramArguments'?

You are right. The KeepAlive key was apparently the problem. But I don't really get it.

twtwtw wrote:

what you're basically doing with it is forcing the shell script to relaunch if it quits with an error.

Isn't that what I want? I want the shell script to be run again, i.e. rsync would attempt again to finish the backup, until rsync quits without an error, i.e. the backup is done.

But actually, this is not what is happening. I sometimes do get an error from the rsync script but the script is not run again. Interestingly, the error only shows up in the rsync log file, but not in the error file I defined in launchd:

          <key>StandardErrorPath</key>
          <string>/Users/marcus/tmp/launchd_pix.err</string>

But again back to the matter at hand. I cannot confirm your guess that

1. launchd runs the job and launches the shell script
2. the shell script launches rsync and quits
3. launchd sees the shell script quit and kills all its child process (e.g. rsync)

Actually, rsync does most of the times complete the backup successfully and is not killed by launchd.

But I don't really understand what the KeepAliveKey has to do with the fact that the script was run at every start of the Mac, but also at the designeted schedule.

And how do I now make sure that the script is run again until rsync exits without error?

Hi VikingOSX. Thanks for these explanaitions, that is helpful. As it turns out, it was the KeepAlive key rather than the RunAtLoad key that made the trick. In any case, if I understand it correctly, the default value of RunAtLoad is 'false' in any case.

Hey twtwtw. Thanks so much for your effort. I have seen many discussions in the forum talking about the problem of launchd restarting processes every 10 seconds or so. In that case, your analysis might be correct. In my case, however, this is not happening. So I cannot confirm your flow, at least not for my case

twtwtw wrote:

1. at the correct time, launchd runs the job, which in turn launches the shell script
2. the shell script launches rsync and then ends
3. rsync starts trying to do a backup
4. launchd sees that main shell script ended and kills all its child processes
5. rsync is killed ungracefully by the system
6. because it had to kill child processes, launchd marks the job as having an unsuccessful exit
7. launchd launches the shell script again immediately, because of the KeepAlive options you've chosen.
8. repeat steps 2 through 7 every ten seconds or so until the machine is restarted (or rsync quits gracefully)
9. machine restarts, but launchd still remembers if the last run of that process ended unsuccessfully and tries to start it again.

In my case, rsync runs, either sucessuflly or with a connection error (usually a time-out problem with the server) and then quits gracefully. It is not killed by launchd. BUT it is also not restarted by launchd if an error occurs.

That means, I guess:

1. at startup or at the correct time, launchd runs the job, which in turn launches the shell script
2. the shell script launches rsync and then ends
3. rsync starts trying to do a backup
4. launchd sees that main shell script ended and hence the keep alive value is fulfilled and it does not restart the script, even if rsync quits with an error

As you suggested, I have a relatively easy script that just runs a backup in rsync:

#!/bin/sh

SRC="/Volumes/Data/Pictures"

DEST="rsync@cloud9.local:/volume1/Backups/"

OPTIONS="-e ssh -E -azvhHS --delete --numeric-ids --log-file=/Users/marcus/tmp/rlog_pix"

EXCLUSIONS="--exclude .DS_Store --exclude */.Trash --exclude .Spotlight-*/ --exclude */$RECYCLE.BIN"

/usr/local/bin/rsync $OPTIONS $SRC $DEST

exit 0

So, as the script is not run continuously when the Mac is on but just once at login (plus at the scheduled times), I still don't understand why KeepAlive was the reason that it run at startup.

Anyway, my script now looks like this:

<?xml version="1.0" encoding="UTF-8"?>

<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">

<plist version="1.0">

<dict>

<key>Label</key>

<string>org.jenal.picture_backup</string>

<key>ProgramArguments</key>

<array>

<string>/Users/marcus/bin/enterprise_pictures_rsync.sh</string>

</array>

<key>StartCalendarInterval</key>

<dict>

<key>Hour</key>

<integer>11</integer>

<key>Minute</key>

<integer>30</integer>

<key>Weekday</key>

<integer>1</integer>

</dict>

<key>StandardOutPath</key>

<string>/Users/marcus/tmp/launchd_pix.out</string>

<key>StandardErrorPath</key>

<string>/Users/marcus/tmp/launchd_pix.err</string>

</dict>

</plist>

That basically does the trick for me. The only thing that it is not yet working is that launchd restarts the script if rsync quits with an error. I haven't figured out how to do that. I guess the script would have to wait for rsync to complete and then either quit if there was no error or report an error to launchd if there was so it could be restarted. Or the restart if error could also be done in the script itself, I guess.

twtwtw, thanks so much for that, this is awesome. Now I also understand what is happening and after all it has a certain logic to it. But I'd probably never have found out. I will try your script and let you know if it works.

twtwtw, I tested your script and it works like a charm. Thanks. I included the exclusions (through --exclude-from and linking that to a text file with the exclusions).

It does not start at every start and it does keep the script alive when rsync quits with an error.

Here is the final launchd code:

<?xml version="1.0" encoding="UTF-8"?>

<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">

<plist version="1.0">

<dict>

<key>Label</key>

<string>org.jenal.picture_backup</string>

<key>Program</key>

<string>/usr/local/bin/rsync</string>

<key>ProgramArguments</key>

<array>

<string>rsync</string>

<string>-e</string>

<string>ssh</string>

<string>-E</string>

<string>-azvhHS</string>

<string>--delete</string>

<string>--numeric-ids</string>

<string>--exclude-from</string>

<string>/Users/marcus/tmp/exclude.txt</string>

<string>--log-file=/Users/marcus/tmp/rlog_pix</string>

<string>/Volumes/Data/Pictures</string>

<string>rsync@cloud9.local:/volume1/Backups/</string>

</array>

<key>StartCalendarInterval</key>

<dict>

<key>Hour</key>

<integer>11</integer>

<key>Minute</key>

<integer>30</integer>

<key>Weekday</key>

<integer>1</integer>

</dict>

<key>KeepAlive</key>

<dict>

<key>SuccessfulExit</key>

<false/>

<key>AfterInitialDemand</key>

<true/>

</dict>

<key>StandardOutPath</key>

<string>/Users/marcus/tmp/launchd_pix.out</string>

<key>StandardErrorPath</key>

<string>/Users/marcus/tmp/launchd_pix.err</string>

</dict>

</plist>

Hi twtwtw. I was a bit hasty with my appreciation. Good thing is launchd does not start the script at every start of the Mac, but it does not start the scripts at the given schedule either. When I start manually (terminal: launchctl start org.jenal.picture_backup) it works fine. But the schedule does not work. Do you think this is because of the AfterInitialDemand key?