Change Polling Time Of Devices To 10 Seconds

Question

On our solarwinds server we are currently monitoring 800 devices (Roughly 500 Network via snmp, 300 Windows via agent)

The polling rate for all network devices is set to every 120 seconds.

I have been asked to change the polling interval from 120 to every 10 seconds for 350 network devices.  The reason for this is to get more accurate stats,  It it believed solarwinds is not accurately capturing ping loss and latency.

What are your views on this

rfroembgen · Answer

Hi jbarr3tt​,

in addition to deverts​ answer i would like to add that solarwinds already does something like you want to achieve manually.

If we take the 120 second default polling interval for latency / packet loss metrics and one of the polls fail, solarwinds will retry polling the target device. If this one also fails, solarwinds will enter the "fast polling" mode for this node. This means that every 10 seconds it will try to poll the node again, as long as it is not reachable.

An example:

First poll completes - next poll in 120 seconds

Second poll failes - Node still be marked as "up" - next poll as soon as possible(!)

Szenario one:

Third poll completes - node status still be up - poll is also counted as up/successful - next poll in 120 seconds

Szenario two:

third poll failes - entering "fast polling mode" for this node - node goes into warning state - next poll in 10 seconds

polls continue to fail - after failing for 11 more polls node goes from warning to down state

solarwinds continues the fast polling untill the node is up again.

However these "fast polling" metrics does not count into availability and also are not stored in the database.

So if the initial poll fails and also lets say 4 fast-polls, but the 5th fast poll will success, the 5th poll data will be stored in the database, but fast-poll 1 to 4 will not.

My recommendation is not to put the 350 devices down to a polling interval of 10 seconds. You could look at this https://thwack.solarwinds.com/message/142870#142870 post from Karlo.Zatylny​ to increase the retry counts after the initial poll failed.

Hope this helps.

Regards

Rene

deverts · Answer

jbarr3tt​,

I've had this conversation a number of times and always with the same type of person...someone that doesn't understand networking.

#1: ICMP - this is a tool for troubleshooting only, and is considered by all devices as a low priority for processing. So, if the device is busy processing real workload, it may or may not respond to your ping (or other management requests (like SNMP) for that matter). For that matter, if any device along the path is busy, it may not respond. So, while ping is great "indicator," it is not without it's shortcomings.

#2: Latency - this is simple physics. Latency is how low it takes to get from one place to another...it's based on distance only. Roundtrip is VERY often confused with latency by someone that doesn't understand the difference.

So, if your 350 devices are having problems and dropping "pings" or are having "latency" problems, there will be other indicators; like link errors, CRCs, or the servers are busy doing workload tasks. Changing the polling to 10 seconds will not solve the problem, and may or may not even reflect the problem. I've only seen someone change the polling rate to be more granular for 1 reason, and that's for bill back metrics. Otherwise, the defaults are fine for 95% of us.

D