Second Polling Engine

Hello everybody,

I am have installed a second SLX polling engine to relieve the stress on our over-burdened original poller. I've run into a curious discrepancy in the Polls per Second tuning wizard. Both servers are identical VMs.

On the original server, recommended PPS Tuning:

Node and Interface Status Polls: 241/sec

Statistics Collection: 98/sec

On new server, recommended PPS tuning:

Node and Interface Status Polls: 8/sec

Statistics Collection: 8/sec

What reason would there be such a large difference between identical servers? If I ignore these recommendations, am I in peril of a poorly performing secondary poller?

Any help would be great!

Thanks, Dan.

Find more posts tagged with

Accepted answers

All comments

njoylif

That is based on the number of Nodes/interfaces and frequency that the poller has to poll.
I would guess you have not put many devices on the new poller yet.
If you want to even out some, you will need to move about half of your pollers (based on nodes/interfaces/poll frequency) over to the new system. Then take a look at that value.

BTW, Love the Avitar

warbird

njoylif is correct. I run 4 pollers, including the primary. We are monitoring 1100+ nodes and ~20,000 elements. As I brought on each new poller, I moved a bunch of nodes over to it. With each new set of nodes, the recommended numbers in the PPS utility increased.

If you are unaware, an easy way to move nodes to your new engine is to fire up the Monitor Polling Engines tool on your primary poller. In the top tool bar is an icon for Poller Load Balancing. You will need to have shut down your polling engines before moving nodes over. Once you've moved a bunch of nodes over, rerun your PPS utility on the new poller. I forget if you need to let the poller run for a few before you run PPS again or not.

HTH.

Mr._Wick

Njoylif - You are correct, I hadn't yet moved any devices to the second poller at the time I posted. I just assumed that the recommended numbers would be based rather on specs rather than load. You know what that gets you. After moving about a quarter of my elements, the numbers were indeed completely different.

Warbird - I used the Load Balancing feature. Worked pretty slick, but not quite as advertised. It's pretty geeky for me to be excited about this, but I am.

Thanks both of you for the quick response and help!