Polling server setup for MPE and APEs - best practice question

Grettings All and Happy Holidays.

Working in an environment with one Main Polling Engine (MPE) and several Additional Polling Engines (APE). SW Agent used throughout for polling.

Question about best practice for polling setup of polling engines (under Node >> Edit Properties).

I understand that best practice will be "APEs polled by the MPE".

But, what to set up for what polling engine polls the MPE?

Can I choose any of the APEs? Or should the MPE poll itself? Or other.

Thanks in advance for any input from the community.

Find more posts tagged with

polling method

mpe

ape

Accepted answers

vinay.by

There is no definite answer for this, ideally yes we poll APEs from MPE and we can choose any APE to poll MPE (but my suggestion would be to use an APE that is less loaded).

In the past if you remember we had few free tools which we would ideally use for self monitoring, or we would use a tool outside of SolarWinds to monitor our SolarWinds infrastructure as part of best practices, but in an evolving world with optimized setup your queries are valid where we do self monitoring within our box.

Hope it helps, happy holidays.

bobmarley

An additional suggestion. If you own HCO and have the free lab license its not a bad idea to monitor your prod Solarwinds environment from your lab as a second set of eyes. It's free and exercises your lab environment.

HerrDoktor

If your MPE is down, there is no Monitoring. Technically you should at least have the MPE in a HA Pool. If this is not possible, I usually opt for an external tool that’s watching the watcher.

All comments

vinay.by

There is no definite answer for this, ideally yes we poll APEs from MPE and we can choose any APE to poll MPE (but my suggestion would be to use an APE that is less loaded).

Hope it helps, happy holidays.

robert777

Thanks @vinay.by !

Yes, I played with this a bit. I have APE1-x servers set up to poll from MPE1 (made-up name) and the APEs poll fine.

When I set up MPE1 to poll itself then it, and all the APEs, went offline (we use Agent for status). "Offline" means status ball red; no effect on polling server function.

So, set MPE1 to be polled from my most lightly used APE and everything seems to work and poll fine.

Marking your solution as verified answer. Thanks for your help.

bobmarley

Had never given that any thought, always had mine in a circle where each monitored the next one until the last one went back to the first one. The MPE ended up on APE1, just by chance.

bobmarley

HerrDoktor

If your MPE is down, there is no Monitoring. Technically you should at least have the MPE in a HA Pool. If this is not possible, I usually opt for an external tool that’s watching the watcher.

robert777

Thank you @vinay.by , @bobmarley , @HerrDoktor . All great advice.

A few things that stand out to me from the thread.

Poll polling servers from a different polling server.
No specific architecture required for polling server polling. Common designs seen in the field: a) star topology with APEs polled by the MPE, MPE polled by any APE (in that site, APE with lower workload is better); b) ring topology, where poller1 polls poller2, poller2 polls poller3, etc.
If your MPE goes down, Solarwinds monitoring stops. Consider a) monitor your polling engines, especially the MPE, with an external monitoring solution (like a Solarwinds lab installation if you have the license), b) spend the money to set up your MPE for HA.

Anyone in the community feel free to question or comment.

brscott

Ultimately (as a couple have already mentioned) external monitoring is the only answer.

What and how you monitor is going to depend on a lot. Yeah there are some basics and someone should (if they have not already) post an article about it, such as monitoring the main URL, maybe hitting this url with an swql and checking the results ( https://[your url here]/Orion/Admin/swis.aspx ), using the OrionAPI to check on things, direct database queries, etc...

What I want to mention is that if you have a ticketing system that you can setup a heartbeat system. I created a special node "Orion-ServiceNow-Heartbeat" and gave it the IP address 127.0.0.1. There's other ways to handle it, but I liked this one. I created an alert that goes off on that node and I didn't include the status of the node, so the alert will trigger no matter what. The alert automatically clears after a few minutes so it can go off again and again, constantly 24 hours a day. I don't know the servicenow specifics but the concept is that it mostly ignores this specific alert but keeps track of the last time it saw this alert come into the system. It then monitors that timestamp to see if it gets past some threshold. If it's been too long it creates a ticket about the solarwinds heartbeat not reporting in.

As a side note: The worst and best thing to monitor the system is humans. If your helpdesk is large enough, you will have that one really annoying operator that is always calling you to let you know something is not right. You can choose to yell at them like everyone else about calling you on "stupid" stuff... OR... you can take the time to research the issue and explain it to them. 90% of the time it will be a waste of your time, 9% of the time you will learn something in the process of researching it, and that 1% they might actually stumble on a problem that it unnoticeable to automated monitoring. Even if that last number is 0.01% of the time... if your monitoring is a major part of keeping everything running smooth, then it can be worth it.