I think that is the general consensus. Trapping at first glance appears to be the solution for alerting until the reliability comes into question due to the lack of some sort of ACK within the protocol. Forrest has talked about this numerous times and likely has some hard fought lessons learned. Maybe he will chime in.
It would be nice if Cacti had some sort of statefull mechanism to switch to a higher frequency “ping” once a device fails to respond to the standard 5 minute check-in schedule. A 5 minute old alert would be fine but you really need to wait for two or more poles to confirm and with such a slow sample rate 10 or 15 minutes becomes a little tardy for some targets. The other desirable requirement is for a hierarchal tree so that a backhaul outage doesn’t trigger a landslide of alerts from all the devices on the other side on the downed link. I believe there are other monitoring systems that do both of these things but I have no experience beyond Cacti. PC Blaze Broadband From: Af [mailto:[email protected]] On Behalf Of Chuck McCown via Af Sent: Monday, December 29, 2014 12:18 PM To: [email protected] Subject: Re: [AFMUG] traps vs polling If we poll every minute or every 5 minutes, catching traps may not be important. If we don’t poll, I worry that we will miss a trap due to it being UDP. From: Eric Muehleisen via Af <mailto:[email protected]> Sent: Monday, December 29, 2014 9:52 AM To: [email protected] Subject: Re: [AFMUG] traps vs polling Both. Traps for realtime alerting and polling for historicals. On Monday, December 29, 2014, Chuck McCown via Af <[email protected]> wrote: When attempting to collect system wide status and alarm data real-time, are traps reliable enough or should you just poll everything on a regular basis? -- Sent via mobile
