Data Getting missed from polls

We have deployed Solarwinds NPM 8.5 SLX with SP3 for our country level Network Monitoring which is connected via MPLS cloud. I have more than 550 devices deployed this contributes to more than 6000 interfaces (elements).

For this deployment we have used a server with following specification
IBM X seris 225
Intel xeon 2.8 Ghz
2.5 GB of RAM
300 GB of Harddisk space.

This is a dedicated server with windows 2003 Server Standard edition.
I am using default MSDE as my database server on the same hardware which has a current database size of 2.8 GB

Right from the begining i am experiencing the problem of data getting missed or not polled from various devices and sometimes few interfaces of a same device. This also includes the network nodes which are connected on the same LAN of the datacenter so there is no issue in the MPLS cloud. Server resources are also not highly utilized.

When i checked the same device on my old monitoring software it shows the correct and complete data so the problem is not with the device but with the orion NPM.

Can anyone suggest what could be the reason and how to reslove the same. We are using this product for SLA monitoring so that makes the database too critical to be lost. It is also next to impossible to get all the 550+ devices reconfigured in the tool again.

Regards

Rajiv

Find more posts tagged with

8.5

Accepted answers

All comments

savell

Have you performed any poller tuning Rajiv? From my experience, 6000 elements on a single server will require the number of pollers to be increased to avoid missed polls.

We generally try to stay within 5-6000 elements per poller. The provided Poller tuning tool suggests 80 pollers at this size, but we have found better results doubling this to 160. I'm guessing however that a number of factors will influence this for each implementation (i.e. this may not be the same for you)

Also, I suspect that using MSDE that you will quickly run out of database space (doesn't MSDE 2005 have a 4Gig size limit?).

Sav.

rajivbahl

Thanks Sav

Definately MSDE has 4GB of limit i am already planning to migrate the complete database to SQL 2005 standard database for which i have raised a request to my procurement team. But that will take atleast a month more. my current database size is 2.4 GB only 60% of the limit.

I have logged a case with the solarwinds team also we checked by increasing the poller tunning value from the default 30 to the recommended 140 but that made the situation bad , polling completion value dipped down to 98.83 from 99.43. So we have to roll back to 30

To be precise i have 541 nodes and 6267 interfaces.

aLTeReGo

Using perfmon on the Orion server, what's your average disk queue length?

rajivbahl

Can you please let me know how to find that out.

qle

One way to get to it is Control Panel -> Administrative Tools -> Performance. Once it's opened, there are three counters that are being monitored, one of which is "Avg. Disk Queue Length".

rajivbahl

Thank for the info

On the Database server the average Disk Queue Length is less than 10 . % Processor time is Average 5.

On the Poller Server the Disk Queue Length is less than 5 and the % Processor time is Average 45.

I hope this is under controlled condition. I have RAID 5 configured on both the servers.

Regards.

Rajiv