This setting is in the web UI under: Admin -> Thresholds & Polling -> Orion Polling Settings
The detailed explanation I find is:
"Default Node Topology Poll Interval - Topology statistics for nodes are regularly polled every configured time interval. By default, this interval is 30 minutes."
I know most people read this and immediately understood at great depth what NPM is doing and moved on with their lives, but I would love a smidge more of an explanation.
Question 1: What is "topology" in terms of what data NPM collects? I saw some posts on Thwack that indicated it was related to "network topology", and then some that indicated "topology of nodes" and making this sound like it was tied to the "list resources" scan you can do on a node. Is "topology" determining connections between nodes (like network topology) or connections between components inside a node (like using the "list resources" button on a node), or both? Basically: Is there a way to determine what devices/nodes/etc. are being polled for "topology". Can I turn this off at a per node level?
Question 2: Can I look at the "discovered topology" data somewhere?
Question 3: By default its set to 30 minutes (globally, per the setting above). If it's network *or* node level topology, this seems a bit extreme unless your network is changing frequently. My Default Node Topology Poll Interval is currently set to 24 hours, which support pleaded
with me to set back to at least 12 hours, I assume because it was so far off from the default. Is there a good reason (other than to keep "topology" data more up to date) to have it set low (like 30 minutes) or some system destabilizing feature that may arise if it's set too high (like 99 hours)? I saw posts that said "be careful setting it lower as it can freeze your NPM system" which makes sense if it is heavy on the poller resource usage, but no recommendations that said a higher value was undesirable ("if set too high, NPM may experience an existential crisis and refuse to poll").
Question 4: I saw some older thwack posts indicating their poller CPU usage went *way down* after turning topology polling off. Has anyone seen topology polling still have these types of issues under certain circumstances?