Is there any way to limit how many probes are broadcast at one time? On my server I have over 1000 services that are probed and the DUDE probes them all at once. This causes my cpu usage to spike to 100% and then I get all kinds off false failures.
Ideally I would like to limit the amount of active probes to 10 or 20 at a time before it sends out the next block.
I know the easy solution is to get a better server but this small tweak would allow the dude to run better on lighter weight machines as effectively as my main server.
edit: I did find a cumbersome way of accomplishing this. If you manually re-probe blocks of devices at different times it seems to stager when they next send out there scheduled probes. This is a nightmare to have to do every time the DUDE is restarted and you have no way of seeing which one is probing when.
edit: The machine I have it running on is a celeron D 2.4 with 1M mem and 40G HD running routerOS 5.0rc3 demo licence. However my main server is a dual xeon 3G with 2G mem and 100G HD with Windows XP and it also runs into this issue but the CPU doesn’t peg out to 100% just 40% when the probes are running. The probe interval is set to 2 min on both machines.
As an update: I downgraded the server to routerOS 4.13 and the cpu usage is worse. It now stays at a constant 100% and it looks like there may be a memory leak.
Ok, I bit the bullet and moved to a xeon dual core and now my cpu is not maxed out anymore. I do however still see a lot of dropped probes and it feels like it is because of how many are running at once.
I use system internals process explorer to force the dude server to run “real time”.
This cuts down on false positives but doesn’t entirely fix them. I also set the 2nd dude process to real time as well (it is the web server so it doesn’t matter).
I also re-probe false positive items to clear them but I have configured all my probes to have a 29 second negative cache, any outage will last 300 seconds if you don’t modify negative cache (on each probe, yuk).
Can you set the processes priority in routerOS? I usually forget to change the priority on the windows server when I reboot. I have been looking for a way to make it permanent but the Google searches Ive done where vague on doing this.
I was just reading up on some of your posts about negative cache and I think that will help some. Just not looking forward to editing all those probes on two different servers
I don’t know about RouterOS. There are applications that will force a service to a set priority at startup. I was going to go find one again but have not gotten around to it.
