doc/79-Jobs.md: one more section

doc/79-Jobs.md

@@ -0,0 +1,77 @@
+Jobs
+====
+
+Director allows you to schedule eventually long-running tasks so that they
+can run in the background. Currently this includes:
+
+* Import runs
+* Sync runs
+* Housekeeping tasks
+* Config rendering and deployment
+
+This component is internally provided as a Hook. This allows other Icinga
+Web 2 modules to benefit from the Job Runner by providing their very own Job
+implementations.
+
+Theory of operation
+-------------------
+
+Jobs are configured via the Web frontend. You can create multiple definitions
+for the very same Job. Every single job will run with a configurable interval.
+Please do not expect this to behave like a scheduler or a cron daemon. Jobs
+are currently not executed in parallel. Therefore if one job takes longer, it
+might have an influence on the scheduling of other jobs.
+
+Some of you might want actions like automated config deployment not to be
+executed all around the clock. That's why you have the possibility to assign
+time periods to your jobs. Choose an Icinga timeperiod, the job will only be
+executed within that period.
+
+Execution methods
+-----------------
+
+Jobs are executed on CLI, basically with the `jobs` CLI command and it's
+available actions and options. A call to...
+
+```sh
+icingacli director jobs run
+```
+
+...would run all currently pending jobs just once. As cron jobs should not run
+forever, the command terminates after this. In case you want it to keep on
+running, just add `forever` (or `--forever`) to the command:
+
+```sh
+icingacli director jobs run forever
+```
+
+This could be used to run the Job Runner as a daemon, preferrably with a
+systemd unit file looking as follows:
+
+```ini
+[Unit]
+Description=Director Job runner
+
+[Service]
+Type=simple
+ExecStart=/usr/bin/icingacli director jobs run forever
+Restart=on-success
+```
+
+However, `forever` is not forever. In case Director detects that too much
+memory has been used (and not freed), it terminates itself with exit code 0.
+So, like the above init script, please expect the Job Runner to terminate at
+any time.
+
+Want so see more details? Add `--verbose` to get colorful log lines on STDERR.
+In case the Job Runner is running with Systemd, those log lines will find it's
+way to your system log.
+
+Time periods
+------------
+
+Icinga time periods can get pretty complex. You configure them with Director,
+but until now it didn't have the necessity to "understand" them. This of course
+changed with Time Period support in our Job Runner. Director will try to fully
+"understand" periods in future, but right now it is only capable to interpret
+a limited subset of timeperiod range definitions.