Traffic queueing on Proxy and DLP hitting 100% CPU when using shared resources

Products

• Data Loss Prevention

Issue/Introduction

When using shared resource pool, you see DLP hitting 100% CPU usage and restarting. 

Environment

Release : 15.8

Component : Network Prevent Email/ICAP

Shared Resource Pool

Cause

Shared resource pool will have depredated resources VS physical dedicated resources. In addition, the resource management AI is likely not providing all the expected logical CPU cores that has been configured in DLP based on system logical cores. 

 

With the most common version of shared resource pool AI, a determination of requested CPU is made based on average overall usage over time. This may prevent the resource pool management from giving DLP additional resources during traffic influxes. 

Resolution

Reach out to Virtual (VM) team and confirmed all cores are being provided when needed. DLP only uses hard tuning based on configured message chains and connections.

The way the resource manager works is that it will take an average CPU usage over time to make a decision and what it will give the server hosting DLP, and may not deliver all CPUs when traffic spikes happen. This can be avoided by using dedicated resources for DLP.