I have an ASA 5540 with multiple context mode enabled. My monitoring host running opennms is behind one context, and hosts that it is monitoring are behind other contexts on the same asa. to get to the other contexts, traffic must pass through the dmz, which is another context. What is happening is that the ICMP messages that are sent by the monitoring host get "stuck" in the DMZ context... the DMZ context seems to amplify and loop the echo requests until it is processing about 350,000 icmp packets per second. these packets that get stuck are never seen by any other host or context, but get passed to the ips module, bringing it to 100% cpu.
The packet capture tool in ASDM shows that all the packets are originating from the monitoring host, and they continue to be captured even if the host is turned off.
I think the packets that are getting stuck are the ones that Opennms' auto discovery sends out and don't have a live host associated with them. I don't get any ping failures to active hosts from the monitoring at all.
Has anyone seen this behaviour before? Any ideas as to what would cause this, or how to fix it?