| Author |
Message
|
| vaf |
 Posted: Fri Nov 21, 2003 2:56 pm Post subject: 2059 errors with load test |
 |
|
Newbie
Joined: 21 Nov 2003
Posts: 3
|
We had some exception on the WEB client which was doing a lot of MQ connects/disconnects under heavy load.
So we wrote test program which connects to MQ using client mode, opens a queue, closes the queue and closes the connection.
All these steps in a loop.
MQ5.3.04 and JARs from the same version.
Running client on WinXP, tried pointing to Linux and AIX servers with the same version of MQ and got same error.
After about 3960 iteractions of the loop we get the exception with 2059 error.
com.ibm.mq.MQException: MQJE001: An MQException occurred: Completion Code 2, Reason 2059
MQJE011: Socket connection attempt refused
There seems to be a lot of open sockets on the XP client PC in TIME_WAIT state when this happens.
If the client PC is slowed down (running a compiler at the same time  ) the problem goes away.
The same client test code runs OK on Linux or AIX.
Anybody experienced anything similar?
Last edited by vaf on Mon Nov 24, 2003 9:23 am; edited 1 time in total |
|
| Back to top |
|
 |
| bower5932 |
| Posted: Mon Nov 24, 2003 7:54 am Post subject: |
|
|
Jedi Knight
Joined: 27 Aug 2001
Posts: 3023
Location: Dallas, TX, USA
|
| Can you post part of a "netstat -a" output here that shows the channels in time_wait state? I'd also double-check that you are disconnecting your sessions and setting them to null. There may be something going on with garbage collection that isn't getting the resources completely freed. |
|
| Back to top |
|
|
| vaf |
| Posted: Mon Nov 24, 2003 8:21 am Post subject: |
|
|
Newbie
Joined: 21 Nov 2003
Posts: 3
|
Some of the time waits.
TCP 10.1.10.201:2503 10.1.13.160:1414 TIME_WAIT
TCP 10.1.10.201:2504 10.1.13.160:1414 TIME_WAIT
TCP 10.1.10.201:2505 10.1.13.160:1414 TIME_WAIT
Total number of these lines is around as many as MQ connections that were made successfully. |
|
| Back to top |
|
|
| Intruder1 |
| Posted: Mon Nov 24, 2003 8:32 am Post subject: |
|
|
Newbie
Joined: 05 Nov 2003
Posts: 6
|
Do you see these ports being in time_wait state on the server side as well? On unix you could set tcp_time_wait_interval or tcp_close_wait_interval to be less than default 4 minutes.
I don't know the appropriate variable and value on the XP machine. |
|
| Back to top |
|
|
| vaf |
| Posted: Mon Nov 24, 2003 8:45 am Post subject: |
|
|
Newbie
Joined: 21 Nov 2003
Posts: 3
|
On UNIX these do not accumulate at the same rate. I see less than 100 of these when the error happens. I will try changing timeout it on UNIX but I do not think it will help given the number of sockets in TIME_WAIT state is low. It also tells me the connections are being closed.
We tried adjusting socket timeout value on XP down to 30 seconds, but it did not help. It is a registry value and I do not remember exactly what it was. 30 seconds was the lowest possible value. |
|
| Back to top |
|
|
| dtauzell |
| Posted: Wed Dec 10, 2003 8:20 am Post subject: |
|
|
Apprentice
Joined: 23 May 2002
Posts: 37
|
We had serious issues like this when our server was under heavy load. We fixed this by creating a pool of connections that is shared by the whole appserver. The application had to be changed to checkout from the pool before sending and then return to the pool afterwards.
-Dave |
|
| Back to top |
|
|
|
|
