From: Jason L. <ja...@oa...> - 2006-04-21 14:32:30
|
I'm looking at using jffnms in our network. The interface is great, and it looks very promising. What I'm stumped on is the performance of the poller. To start, I've run autodiscovery against one of our routers, a cisco 6500, and it produces 599 interfaces. poller2.php never completes (it just times out), and poller.php takes > 5 minutes to poll just that one host: 10:20:44 : H 2 : Poller End, Total Time: 539131.16 msec. The machine and network are quite sufficient, I believe. When I run poller.php manually, it looks to be spending most of its time doing the snmp queries. A glance at the code looks like it is running snmpget repeatedly for each OID. Have there been any efforts to rework the poller to not have this (process forking) overhead? Are there others running jffnms' poller with a similar number of interfaces on a single host? If the poller should scale to this size, can anyone recommend where to look for my performance bottleneck? Thanks! -j -- Jason Lavoie Ratvarre sbe uver ja...@oa... |