Whats wrong with SNMP service?

[rweales]

I run Cacti Version 0.8.6i (patched to j) on FC4, not Windows. It works great.

My question is about the SNMP service on 2K3 server.

I monitor several 2K3 server boxes with Cacti, via the SNMP service. The service usually runs fine, however in the last few months, and more so recently, the service chokes and won't run.

I notice that a box isn't graphing, so I run a verbose query and get nothing back. I will check the box, and see that SNMP isn't running. Event log will sometimes show an error, sometimes not; never anything with detail to discern what is causing the problem with the SNMP service. I restart SNMP, get a query, and then the service will die after a few minutes or hours. This has happened on different servers performing several different functions: File servers, Backup servers, Web servers. So the functions of the servers where SNMP dies are different.

Of course, I've never had this problem on *nix...

Once I can't get the service to stay running I uninstall it and reinstall it. Sometimes that fixes the problem, but lately it doesn't. I have had three in the last two weeks that I haven't been able to get SNMP to run consistently on. WTF?

Is there a "neutral" SNMP client for Windows that I can use instead of the MS one that won't run consistently?

Is this just Windows updates breaking it?
Will SP2 fix this?

Any suggestions are appreciated!

[BSOD2600]

What errors are in the event log related to SNMP?
Have you googled them?

I've got several Win2K3 SP1 servers and their SNMP service runs just fine...

Does the server have SP1 installed? If not install that. If you're feeling brave, install SP2. There is a Microsoft KB on exactly what changes/fixes it makes.

Net-SNMP has an agent for Microsoft, but it doesn't fully support everything the Microsoft agent does.

[rweales]

Most of the time event log just says "SNMP service shut down". I have searched quite a bit on this issue, and finally came here thinking someone may have seen this before. What is strange is that it happens on boxes that are used for many different purposes.

I'm going to try SP2.

[eviln1]

Greetings.

I am having the same issues with a couple of XP SP2 (fully updated) boxes... and i dont even have "Service died" messages in the logs.
Most of the time, a manual restart of the service fixes the problem for about 30 minutes... and then it just dies again.
What creeps me out is that randomly, and about once a week, the service "wakes up", responds to a couple of polls and then dies again. I am not sure if the XP box has restarted at that time :s

Tomorrow i will install the net-snmp instead of the XP built-in service, and will let you know how it works out. Frankly i think it is just another windows bug.

[crackrocksteady]

I have seen this when another application is trying to use snmp port and causes conflicts. I saw this with a particular version of the adaptec raid utility.

Have a look on your servers for an application with snmp support that might be binding to that port.

Thanks

Fletcher