Check your throughput. We're field testing a CX module at the latest version of code. We found that at well below the limitation of the 1GB interface the memory locks up and starts dropping packets. At that point, all we could do was power cycle the CX module and until then, it would be an outage.
We were able to license the 10GB port for the demo and memory still runs very high, but we aren't experiencing full traffic stops due to packet loss anymore.
We believe that the CX module has a problem with disregarding old sessions that should be closed, thereby creating it's own denial-of-service attack on itself at lower throughput. It just starts dropping all traffic. Still not sure why the memory allocation goes so high and never lets go. The firewall is usually the bottleneck in the network but the ASA itself is performing to specification. It's just once we punt traffic up to the module for inspection that we lose the device.
At this point, we've been asked to downgrade to an earlier code version to continue field testing.
