Resolving RAM and CPU Spikes

For External Customer / Partner Use Only: These instructions are provided to help customers / partners perform the following action themselves to resolve the issue. If the customer / partner runs into issues following the instructions, please have them open a case with licensing support (https://www.cisco.com/go/scm) to help resolve. Please DO NOT perform these actions yourself if you are an internal Cisco resource outside of the Licensing Support team.

Q: Sometimes the SSM On-Prem’s application RAM and CPU spikes, which causes the application to become very slow. What can I do about it?

A: In our previous releases (before 8-202212), we have observed RAM and CPU spikes caused by certain processes in a couple of scenarios. We have addressed these in Cisco SSM On-Prem Release 8-202212. To aid in resolving this issue, we recommend upgrading to Release 8-202212 or later. You can also mitigate the RAM and CPU spikes by ensuring that you have enough licenses in your SSM On-Prem application so that the traditional SL devices are all in compliance.

We have observed RAM and CPU spikes in certain scenarios such as devices using traditional smart licensing getting continuously bombarded with auth_renew requests from the device to the SSM On-Prem application. To address this, an optimized OK_TRY_AGAIN response has been added. This response is sent to devices when the SSM On-Prem application is loaded with many requests and determined to be too busy in processing the device’s messages. This response notifies the device to retry its request at a certain time interval. As the request load increases, the interval becomes longer to smooth out the request load on the SSM On-Prem server.

Another scenario where we have observed RAM and CPU spikes is during the synchronization of SLP devices. In this scenario the synchronization has not been performed for a very long period, and now the SSM On-Prem application has a large amount of SLP data to sync with CSSM Cloud. To mitigate this, we recommend performing the SLP synchronization with CSSM Cloud, with a load of 10,000 devices, once per week.

Note: For certain processes, such as SL and SLP Synchronization with Cisco, it is normal for the RAM to spike for the runtime of the process. However, it should go back to normal after the process is finished. If you see sustained high RAM or CPU spikes, please contact TAC**.

** Restarting gobackend and puma should help with controlling the RAM and CPU spikes. However, customers cannot perform the restart, so TAC has to get involved.

Troubleshooting:

If you experience an issue with this process, that you cannot address, please open a Licensing case at Support Case Manager (SCM) using Software Licensing option.

For feedback on the content of this document, please submit here .