I have a question about how zenmodeler polls the devices. Is it possible to have an offset to the time that the modeler scans devices?
The reason I'm asking is because I have a setup with right at 100 Windows servers being monitored and every day starting at about 9am zenmodeler kicks off and brings the host server to a crawl with max ram and cpu usage. For about an hour solid I start getting zenmodeler heartbeat errors, timeout errors on WMI connects and event log reads, among a few other error messages. After all the devices have been modeled for the day (I have a 24hr delay set) the server calms down, all the alerts completly stop, and the server goes about it's daily monitoring buisness without a peep until 23 hours later when the modeler starts up again. I don't have the strongest server hosting Zen which may be part of the reason, it's on a VM with 3ghz of cpu and 1.5gb of ram allocated, but it seems if the modeler staggered its device scan throughout the day instead of cramming all the scans into an hour long window the server wouldn't get so hammered. -------------------- m2f -------------------- Read this topic online here: http://forums.zenoss.com/viewtopic.php?p=28346#28346 -------------------- m2f -------------------- _______________________________________________ zenoss-users mailing list [email protected] http://lists.zenoss.org/mailman/listinfo/zenoss-users
