Hook documentation?

[cigamit]

What I really meant was - is there any issue with a fairly long-running process hooking into poller_output?

Just remember that the poller_output hook will get called once per host, so if you are using it to launch a new process, it will launch that new process multiple times. Any other major computations it has to do will just slow down your polling slightly, nothing more.

[Howie]

What I really meant was - is there any issue with a fairly long-running process hooking into poller_output?

Just remember that the poller_output hook will get called once per host, so if you are using it to launch a new process, it will launch that new process multiple times. Any other major computations it has to do will just slow down your polling slightly, nothing more.

OK. So I think I actually need a 3-step process:

poller_top: parse all the map configs, decide if any need poller_output data

poller_output: save away the required data in a table somewhere

poller_bottom: draw the maps (which can have a mix of data sources, so they need to be done after rrd updates in a non-boost system).

The easy way of doing that - holding all the map objects in memory between poller_top and poller_bottom - sounds like it'll be quite memory-hungry. and worse, memory hungry before the poller has run, so possible to kill it Time for a rethink.

[rony]

Memory table in mysql 5.x and greater? And a temporary table in 4.x and less?