Hi Mike > Looking at the differences between 1.2 and 1.3 it appears that keepalive > connection optimizations were put in place that perhaps caused this > regression. I say regression, because I don't see the infinite loop when > running 1.2 synapse with the simple telnet 8280 test. Also, I did not > see this error in any of my testing until I deployed to a box which was > subject to keepalive processing. I'm definitely no expert in this area, > so this is all just based on anecdotal evidence - I'm just throwing > these observations out there for you to chew on... > There was a proposal from one of the users that we should recover from fatal exceptions by auto-restarting the IOReactors. I think thats a good thing as certain severe network level failures etc could otherwise put Synapse into a non-recoverable state, in which case shutting down and restarting automatically is better than trying to log the error and continue in an infinite loop.
However, I cannot think why this could happen in 1.3, instead of 1.2 in "normal" processing.. if you could provide a tcp dump (taken on Linux as "sudo tcpdump -i eth0 -X -n -s 4096 port 8280 > dump.txt" would be very valuable to find out the cause of this. Will it be possible to get this log? cheers asankha -- Asankha C. Perera AdroitLogic, http://adroitlogic.org http://esbmagic.blogspot.com
