another admin pointed out something very odd with this particular monitor. apparently, the monitor has some overrides that change the threshold in certain scenarios. to start, the monitor description:
This monitor ensures that the "Process\Handle Count" counter for the HealthService.exe process does not exceed a set threshold over a series of consecutive samples. If the conditions are met this monitor will change to a critical state, which will then roll up to the "Health Service State" monitor. The "Health Service State" monitor is configured to run a recovery when its state is critical, which will automatically attempt to restart the Health Service.
basically once you breach this number, the health service restarts. this is typically a good thing since you’re keeping it maintained. now, flip to the overrides.
notice that there’s an exchange 2007 computer group override where the value is 5000. try to edit this override. you should get a similar screen.
notice how the value of 5000 doesn’t show up here. interesting that it would even be set at 5000 since 6000 would seem a better rounded number for most agents. so why would the exchange 2007 computer group want a lower threshold? mysterious…
not really -- if you know the history. turns out at one point the threshold was set to some whacky low number. i don’t have a back rev environment to go pull the actual value. let’s just say it was 200. with this value in place, the exchange mp couldn’t reliably operate in large-scale environments with health service constantly restarting. the override value comes from the exchange mp, forcing the threshold count to a much higher, more realistic value.
this makes complete sense except the value is lower than what is shown in the screen shot above, right? actually … the value of 6000 was introduced in the latest operations manager 2007 core mp which was released after the exchange mp.
oh by the way, you’ll see this same behavior in the health service private bytes threshold monitor. (thanks guys!)
Comments
Post a Comment