cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
134
Views
2
Helpful
2
Replies

High Memory Utilisation 9200L

alexhunter
Level 1
Level 1

We are seeing random high memory utilisation spikes on our Cisco 9200L switches, running version 17.9.4a. We thought this might be related to the application visibility that is pushed out to them via Cisco Catalyst Centre, so we disabled application visibility across our estate of switches, due to our current environment not giving us any benefit of this being turned on. After disabling it within Cisco Catalyst Centre it does not remove the config from the switches, we are only seeing this random spike on switches with the config left on that was pushed out by Cat Centre.

Looked at the current release notes of IOS XE 17.9.4a and 17.9.5 and there is currently no outstanding caveats related to memory leaks that I can find relating to these releases.

The memory spike is around 74-80% when it happens, where when running under normal conditions it is around 40-50%.

When looking through the memory process utilisation when this happens in the CLI it is always the pubd process being highly utilised but this doesn't necessarily point to what the cause is.

1 Accepted Solution

Accepted Solutions

Leo Laohoo
Hall of Fame
Hall of Fame

Accept the "penalty" (or cost) for the unavoidable memory leak and eventual crash of the switch associated with DNAC.

The process "smand" is attributed to API calls from DNAC.  The more API calls, the faster the memory leak.  "pubd" is another process caused by (or related to) DNAC.   Bug IDs will never use the term "DNAC" or "Catalyst Center" but, instead, use substitutes like "telemetry" as not to upset clients.  And there is a very long list of Bug IDs of memory leaks caused by these two processes.  

Regardless the the version of the firmware, it is best to perform regular (like every 6 months) reboot of the stack.  Cold-Reboot is preferred but warm reboot is better than a crash.  

View solution in original post

2 Replies 2

Leo Laohoo
Hall of Fame
Hall of Fame

Accept the "penalty" (or cost) for the unavoidable memory leak and eventual crash of the switch associated with DNAC.

The process "smand" is attributed to API calls from DNAC.  The more API calls, the faster the memory leak.  "pubd" is another process caused by (or related to) DNAC.   Bug IDs will never use the term "DNAC" or "Catalyst Center" but, instead, use substitutes like "telemetry" as not to upset clients.  And there is a very long list of Bug IDs of memory leaks caused by these two processes.  

Regardless the the version of the firmware, it is best to perform regular (like every 6 months) reboot of the stack.  Cold-Reboot is preferred but warm reboot is better than a crash.  

I had a feeling that may be the case, but I appreciate the info thanks again.

Review Cisco Networking products for a $25 gift card