potenital ITIMER issue with solaris 9
Please help!!. Here is a description of the problem.
I have a E4500 platform with dual processors ( 400 MHZ). I noticed that everytime i run a ftp for a file ~250MB from the E4500 to a host on the other end of the router, i loose OSPF adjacency with my router ( deadinterval is set to 5 seconds). . The links to the router are all FastEthernet full duplex. I am running gated on E4500 and noticed that it is getting CPU starved at certain peaks of ftp traffic for ~20 seconds. I did a truss on gated and confirmed that gated is not being schedule at peak ftp traffic.
Looking closely at truss output of gated i found it uses ITIMERS and expects waking up from the kernel via SIGALRM to send OSPF hello messages. However,the surprising part during the peak ftp traffice this does not happen. Also, the CPU is not maxing out it is at only 30% peak time from "top" output.
I tried playing around with deadinterval and had to set it to 25 seconds in order to not loose adhancency with the router. However, this is a temporary workaround and not acceptable as it impact failover time. I am not sure how gated could be CPU starved for ~ 20 seconds during heavy tcp traffic. I have installed the latest Solaris 9 recommended patches including fixes for hme drivers but no luck. Also, there are no "nocanput" errors".
Any more clues on what could be happenning and how to fix it?