RE: [Quickfix-users] Quickfix/j migration...
Brought to you by:
orenmnero
From: Steve B. <sb...@sm...> - 2006-02-16 09:06:11
|
Hi Chris, What platform (operating system version, Java version, QFJ version. etc.) are you using? =20 > - compilation was straightforward - java code is compatible That's good to hear. > - perhaps obviously, the store files are not compatible, get=20 > EOFException when trying to use JNI quickfix with quickfix/j=20 > - so removed the old files It appears the QFJ files can be used by the JNI code but not the other way around (Unicode issues). If it's important=20 to be able to change libraries during a trading session we could make the files compatible but it's not currently a high priority. > - the quickfix/j code does not seem to like to bind to port=20 > 443 (the jni version did) - it claims the port is in use... -=20 > maybe it's a priviledged port issue, I do have 2 acceptors=20 > running in the same jvm and other one seems to bind to port 442 ok. I just ran the order executor (Windows XP Pro, Java 1.4.2_07, QFJ CVS head) and it had no problem connecting to port 443=20 (after I disabled IIS). > And now to the problem, I am running some performance tests -=20 > using quickfix/j on both sides - a dummy acceptor is pumping=20 > messages into my app and responding appropriately to any=20 > messages it sends. After running for 5-10 minutes, the=20 > initiator then disconnects claiming it is not getting any=20 > heartbeats. I have a setting of 10 (seconds) for the=20 > initiator heartbeat. >=20 >... > I am using the latest beta2 - which I guess should have the=20 > fix to bug #133, re: acceptor sending heartbeats. >=20 > Is the heartbeat behaviour different compared to the jni version? Again, please give me the information on the platform you are using. Can you tell me more about the test you are doing? Is the acceptor=20 sending a constant stream of messages (< 10 second) interval? Is the application (initiator) also responding at a rapid rate? I'm asking because there wouldn't be any heartbeats in that scenario. I've run tests with the CVS head where I sent 1,000,000 messages from an initiator to an acceptor without any heartbeat problems. However, it's possible the problem is in the other direction. Regards, Steve |