Cedric, Thanks for the suggestions, but there are no limits within the network I am monitoring (I design/build/maintain the same network) and we don't use cisco ;) also the routers are sitting at less than 20% CPU util...
Colin -----Original Message----- From: Cedric Charest To: Colin Horsington Cc: '[EMAIL PROTECTED]' Sent: 05/03/2003 5:54 PM Subject: Re: [smokeping-users] Smokeping runnning time - maximum number targets in 5 minutes Well I'm running Smokeping and MRTG on the same server, which is a Athlon 1133 MHz, 256MB RAM. Here, Smokeping is monitoring around ~400-500 elements, and for now, MRTG is monitoring 50 elements (soon around 1500+). The system load was so high that, often, smokeping wasn't receiving ping in time. I fixed the problem with MRTG by puting it as a deamon and by adding 512 new MB in the ram. But this doesn't look to be your case. Also, with someping, I am doing aroung 20 to 50k/s (in and out) of ICMP packets. With 1500 elements that are pooled with 20 pings (84 bytes per packets) every 300 seconds (ie: 5 minutes) is doing a lot of traffic (in packets) of your router/firewall. If your firewall is configured to detect DDOS attacks and try to stop them, it may decide to block pings when there is too much packets. And also, don't forget that icmp packets are the first packets type to be dropped by a router when the router is working too hard... that's why pings aren't really reliable. Hope this may help! ;) Cedric Colin Horsington wrote: >All, > >In trying to get 1500+ hosts working on single smokeping system. After >running smokeping for several days, yet not recieving any output on the >graphs I began measuring the time take to run smokeping *once* across a >given number of probes. There was a thread sometime ago that did cover this >but no clear conclusion was drawn...( >http://www.ee.ethz.ch/~slist/smokeping-users/msg00237.html and >http://www.ee.ethz.ch/~slist/smokeping-users/msg00236.html) > >The server I am runnning is running is a dual 733Mhz with 512Mb RAM with no >signicant load. It is running smokeping 1.18, Linux kernel 2.4.19 and perl >5.6.1. > >Here are the timings for given number of targets... > >#(root)-(1)-(ops1) -=> time /usr/sbin/smokeping --nodemon --debug >### fping seems to report in 1 miliseconds >Launched successfully >FPing: probing 527 targets >/usr/sbin/smokeping --nodemon --debug 1.37s user 0.93s system 1% cpu >3:36.37 total > >#(root)-(0)-(ops1) -=> time /usr/sbin/smokeping --nodemon --debug >### fping seems to report in 1 miliseconds >Launched successfully >FPing: probing 764 targets >/usr/sbin/smokeping --nodemon --debug 1.41s user 0.30s system 0% cpu >5:16.26 total > >#(root)-(0)-(ops1) -=> time /usr/sbin/smokeping --nodemon --debug >### fping seems to report in 1 miliseconds >Launched successfully >FPing: probing 1030 targets >/usr/sbin/smokeping --nodemon --debug 2.06s user 1.68s system 0% cpu >7:22.21 total > >#(root)-(0)-(ops1) -=> time /usr/sbin/smokeping --nodemon --debug >### fping seems to report in 1 miliseconds >Launched successfully >FPing: probing 1279 targets >/usr/sbin/smokeping --nodemon --debug 22.89s user 11.05s system 5% cpu >10:49.68 total > >The times above excludes the generation of the .rrd files as I ran the above >command twice to ensure they were created... > >I have seen some documentation that some probes (that can take only a single >argument) maybe forked for scalability. Does anyone know or have fping >forking. I imagine to do this the number of targets per fping would be >limited (i.e. 100) and you would run a number of these in parallel to >scale...? > >Tobi, thoughts suggestions? > >Thanks, > >Colin Horsington > >-- >Unsubscribe mailto:[EMAIL PROTECTED] >Help mailto:[EMAIL PROTECTED] >Archive http://www.ee.ethz.ch/~slist/smokeping-users >WebAdmin http://www.ee.ethz.ch/~slist/lsg2.cgi > > > > -- Unsubscribe mailto:[EMAIL PROTECTED] Help mailto:[EMAIL PROTECTED] Archive http://www.ee.ethz.ch/~slist/smokeping-users WebAdmin http://www.ee.ethz.ch/~slist/lsg2.cgi
