I need to lower the amount of pollers we currently have, can I unmanage or do I need to delete nodes?

Question

We are in serious need of a NPM Cleanup.  We are polling too many nodes and our NetPerfMonService.exe keeps dying every hour or two.

Here is the Engine Status...any help/tips/suggestions would be great. (fairly new to Orion)

Orion Network Performance Monitor Version 9.5.0
NetPerfMon Engine    
Network Node Elements    2301
Interface Elements    8172
Volume Elements    3085
Date Time    1/10/2010 8:22:41 AM
Paused    False
Max Outstanding Polls    900
Status Pollers    
ICMP Status Polling Index    1292 out of 13558
SNMP Status Polling Index    2762 out of 13558
ICMP Polls per second    100
SNMP Polls per second    0
Max Status Polls Per Second    100
Packet Queues    
DNS Outstanding    43
ICMP Outstanding    884
SNMP Outstanding    1726
Statistics Pollers    
ICMP Statistic Polling Index    4845 out of 22529
SNMP Statistic Polling Index    3486 out of 22529
ICMP Polls per second    65
SNMP Polls per second    0
Max Statistic Polls Per Second    65

rweckesser · Answer

Again thank you for the valuable information.  I will have to look into the null0/loopback interfaces, come to think of it I am sure there are a few being monitored.  Just hard to work with System manager when NPM keeps bombing out.

mrxinu · Answer

No problem at all. Yeah, Solarwinds recommends no more than 8k elements on a poller and you're exceeding that with your interfaces alone. Are you sure you're not monitoring a bunch of loopback & null0 interfaces that could be cleaned up?

You're right on the money with the balancing though. What I'd recommend is creating a custom property called Priority_Level (or something similar) and assigning a priority to all your nodes. Then have someone DB-savvy run a query to update the pollinterval based on the priority level.

Something like:

update nodes set pollinterval = 300 where priority_level = 'P3'

That would set the low-priority nodes (P3) to 300 seconds giving you a little more breathing room. You can make the same kind of adjustment to the statistics gathering, though that value is in minutes, not seconds.

Note: You'll need to check on the field name - pollinterval was off the top of my head. If you have them do a 'select top 5 * from nodes' and look at the values you should see one column with 120 in it and another with 10. That's what you're looking for.

rweckesser · Answer

Thanks for the response.

Also, would reducing the polling interval on say our workstations increase the performance of NPM?  Our netperfmonservice.exe runs at 1.1GB and eventually dies after just an hour or so of polling.

Thanks again.