Showing results for 
Search instead for 
Did you mean: 
Create Post
Level 7

Cisco 3850 Reporting lsmpi_io

We recently installed our first WS-C3850-48P switches. NPM has been reporting high memory utilization, which is attributed to the lsmpi_io reading. I saw that there were similar issues with Cisco ASR devices and a customer poller was created. Are there any pollers out there than can show the correct memory utilization for the 3850's, or an incoming update to the existing MIB? Thanks.

ASR Thread: Memory usage reporting incorrectly on CISCO ISR 4331

Switch readout:

"switch-a" #show memory free

                 Total(b)        Used(b)         Free(b)         Lowest(b)     Largest(b)

Processor 916107072   310302384   605804688   600460920   601617972

lsmpi_io   6295128       6294304       824               824               412

NPM readout:


Labels (1)
0 Kudos
6 Replies
Level 9

I ran into this issue on a pair of Cisco 4451 Gen 3 ISR routers. Same issue with lsmpi_io.  I followed the directions under A Node Shows Incorrect CPU & Memory Utilization - SolarWinds Worldwide, LLC. Help and Support 

In my specific case, I changed the N.Memory.SNMP.CiscoGen4 memory poller to the N.Memory.SNMP.CiscoGen3 poller and they're now showing correct usage values.


Level 7

I ran into the same issue myself on a WS-C3850-12XS-E, and followed the recommendations at this link:

A Node Shows Incorrect CPU & Memory Utilization - SolarWinds Worldwide, LLC. Help and Support

I selected and checked each supported poller until I found one that reported the right results.

Hope this helps.


0 Kudos
Level 7

Sean -

Did you find an appropriate poller to use for your 3850s?  I'm actually running into the same thing.  I don't believe it is reporting on the lsmpi_lo value though.  It is reporting on the output of the following command:

show platform software status control-processor brief

If you take the Used sum / Total sum, you will get the percentage that Solarwinds is reporting.  Currently my Active/Standby switches run at about 84% and my Member switches run at about 51%.  What this leads to is the switch stacks that have more switches show considerably less memory usage than 1 or 2 switch stacks, since the more Member switches in the stack bring down the average.  I'm looking to get something that shows a more consistent memory percentage across the switches stacks that have varying amounts of switches in them.


Brendan, the tldr version is that no, we have not found the right poller. Our switch stack only has three switches in it currently, so it's not maxed out. Has anyone else found something?

0 Kudos

I'm running into similar issues you all describe. Seems the mibs being polled by default are not accurate and i have also tried making my own custom poller using (used) 354916752 (free) 497362912

no luck though as the numbers do not come close to show platform resources.

Any help is much appreciated.



0 Kudos

I found the mibs which are tables and then created custom pollers - mem used - mem free

Manipulate as you see fit.  Did the trick for us.




0 Kudos