Hi, I'm looking at using notifications along with Icinga to avoid running a lot of NRPE instances for doing basic disk-space-free / cpu-load style checks. My employer currently uses Munin to do this.
What I currently do, as per the Icinga documentation, is define a staleness check. If a period of time passes with no passive check being sent then the status is updated to a warning with a description of "stale". The munin-limits command will send a status each time it runs, either OK, CRITICAL or WARNING, but Collectd appears to only do this on a state change. I can see how I can set a plugin as interesting, and therefore get multiple notifications when it's in a failing state, but not if it's in a okay one. I'm considering writing a patch to make notifications happen regardless of state, though perhaps not every <Interval> (that would be quite noisy). Either I'd send it every <configurable factor> * <Interval> seconds or (less pleasant) from a background thread. If I were to do this work to suitable quality, is that in line with where the project is headed (would it be likely to merged into 5) or is this not following the spirit of how you see notifications being used? Thanks in advance, Aaron _______________________________________________ collectd mailing list [email protected] http://mailman.verplant.org/listinfo/collectd
