MikroTik

I'm getting massive memory leakage in all versions I've tried (which is 3.1 onwards... none from earlier). Have tried running as a service, and with remote and web servers disabled - seems to make very little difference. We don't use AutoDiscover. Now we're at a loss - we've put 1Gb of ram in the machine and TheDude just eats its way thru it inside of a few hours. Any suggestions?

As screencaps below show, memuseage is increasing by 200Mb over an 11minute time period...
Once mem useage hits 750Mb it offloads to swap file, and starts again from 30Mb. Hits 750Mb again, offloads to swap, til there is no more swap space, which causes windows to display the good old "Virtual Memory Too Low - Windows is increasing your swap file..." balloontip.

I am running w2k3sp2 and the dude 3.4, don't autodiscover but I only run a few services. Mostly ping on 1600 computers and cisco cpu on 100ish devices. You could check some of your probes to see if that is causing it. Are you running any external commands? I found it to be more stable by increasing the mac mapping refresh interval to 1 hour...Main settings -> misc.

I have found that setting affinity to a single core is good as well. Search for affinity to find the thread about graphing discrpancies and affinity.

I checked over the last 30 minutes and my memory utilization is just below 100mb and actually dropped by 100k. Most of the things I have tried have dealt with graphing issues and never looked for a memory leak before.

(edit add more detail)

My guess is that somehow the dude has been screwed up. It happened to me twice during the many dude 3.0rcX upgrades that I went through. How it is caused, I do not know - maybe I left an old dude running, maybe Windows just broke the dude, maybe something went wrong when I made Dude a service by ticking the box.

My fix was basically reinstall the Dude

1) Shut down Dude
2) Use Task Manager to make sure all dude.exe are gone
3) Copy the "data" directory (usually c:\Program Files\Dude\Data) somewhere else
4) Uninstall Dude
5) Find anything left under c:\program files\Dude and delete it. Make sure no other dude.exe can be found on Hard Disk.
6) Check Dude windows service is gone
7) Reinstall Dude and set it up as a windows service and test (add a device, check is working, etc)
Stop Dude client and service again (making sure all dude.exe are gone using Task Manager)
9) Copy*the contents of your backed up data directory back into c:\Program Files\Dude\data
10) Start Dude service - should be 2x dude.exe at this point, both running as SYSTEM, one using CPU, one not.
11) Run Dude client and connect to service - should be 3x dude.exe as this point 2x SYSTEM; 1x ITTeam
12) Hopefully everything is running well now, and the CPU using SYSTEM dude.exe is not growing.

*Note: Copy data directory. Best to keep a copy in case of a screwup.

I have ~70 devices and about ~700 monitors. My Dude Server (running 3.4; and 3.2 before that) has sat on 50% memory used for months. The only change occurs when I restart the server or upgrade Dude.

mcoyles, can you send your Dude export file to support please? we will look at what's causing this

@sweetdude - nope, no external services etc... just using inbuilt probes - ping usually, and snmp on all links and switches.

@adamd292 - yep that's my usual recovery procedure. Not gone that far yet - figured better to try and work out why it's happening in case it's an issue that needs plugging by Microtik...

@normis - On it's way to you...! 4Mb rar'd

friends,

I am with this problem, and have that be restarting the service. x)

Finally gotten to the bottom of this problem... don't snmp-monitor any PCs / Servers. Restarting the service was not a good enough solution for us... and reinstalls of any version all resulted in the memory leak resurfacing as soon as I pulled our map in (ie: restored our previous data directory).

I binned our map as shown above and started from scratch. Uninstalled TheDude, performed a clean install, and used autodiscover. As soon as it had completed, the memory leak was back again. Next step: work my way thru the map deleting items until it stopped. First thing I did was remove ALL SNMP links... memory leak still present.

As soon as I removed the PCs / Servers that were being SNMP monitored for CPU / Mem / Disk etc, the leak stopped.

I next ran autodiscover again and let it find the entire network and all systems, and then manually disabled snmp monitoring on those computers on which it had discovered it, and sure enough, memory leak stops.

Hope this helps some of you experiencing similar probs!!

Relly, this is my problem...

Thanks mcoyles The above post fixed the problem we were having.
We are currently running just over 700 individual probes.
Found quite afew forum topics mentioning the problem, none apart from this one with a solution.

I found the dude hanging every hour or so, then requiing the service to be restarted "very annoying"
Our Dude was growing to over 1.5GB of memory growing at a rate when the leak starts of over 100Kb/sec.

After reading this changed all servers to the No SNMP profile within the settings page for SNMP monitored servers. I also went ot the trouble of removing the [device_performance()] snmp string from the appearance label (probably overkill).

Since then 100% uptime for the service and more importantly no more gaps in my graphs
The Dude is now sitting at 84Mb a far cry from the 1.5 Gb it was getting around to using previously.

Very very happy with The Dude
__________________
Network Administrator