| Author |
Message
|
| einpajo |
 Posted: Wed Sep 26, 2001 6:07 am Post subject: |
 |
|
Newbie
Joined: 25 Sep 2001
Posts: 1
Location: Sweden
|
We have an application running on Solaris 8 with MQ 5.2. After sending about 7000 small messages the client application hangs and returns error 2059. Our application does connect->open->put->close->disconnect for each message. If we change this so that it does connect->open->put,put,put,...,put->close->disconnect it will work but unfortunately we are unable to use this "method" in our application. We have tried to "tweak" kernel parameters according to tips from this site (when Oracle and MQ is residing on the same machine) but it didn't help.
We also tried to add a tuningparameter to qm.ini without any success:
TuningParameters:
IPCCECSetSize=1000000
Thankfull for help!
|
|
| Back to top |
|
 |
| kolban |
| Posted: Wed Sep 26, 2001 1:53 pm Post subject: |
|
|
Grand Master
Joined: 22 May 2001
Posts: 1072
Location: Fort Worth, TX, USA
|
Sounds like we need to look at some debugging. When you say your client app, do you really mean an MQSeries client application (i.e. one linked with the MQSeries client libraries and can be located on a different machine than that of the queue manager?) or do you mean an app that is a "client" of MQSeries services?
In the queue manager /var/mqm/errors and /var/mqm/qmgrs//errors directory there are error files. Look at these and also associated .FDC files. Any further clues?
Let us know what you find. From here, we will be able to suggest additional tests to be executed. |
|
| Back to top |
|
|
| eindano |
| Posted: Thu Sep 27, 2001 12:48 am Post subject: |
|
|
Newbie
Joined: 26 Sep 2001
Posts: 3
Location: Sweden
|
Yes we mean a MQSeries client application that can be located on another machine, but in our case the queue manager is located on the same machine.
We use MQCONNX, MQOPEN, MQPUT, MQCLOSE and MQDISC calls for every message we write the queue.
In the error log file /var/mqm/errors:
AMQ6119: An internal MQSeries error has occurred ('unknown value of errno' from
shmat.)
EXPLANATION:
MQSeries detected an unexpected error when calling the operating system. The
MQSeries error recording routine has been called.
-------------------------------------------------------------------------------
AMQ6184: An internal MQSeries error has occurred on queue manager TEST
EXPLANATION:
An error has been detected, and the MQSeries error recording routine has been
called. The failing process is process 25606.
And the .FDC file:
Date/Time :- Thursday September 27 09:35:24 GMT 2001
Host Name :- (SunOS 5.
PIDS :- 5765B75
LVLS :-520
Product Long Name :- MQSeries for Sun Solaris 2 (Sparc)
Vendor :- IBM
Probe Id :xY129018
Application Name :-MQM
Component :- xstConnectExtent
Build Date :- May 22 2001
CMVC level :- p520-CSD01G
Build Type :- IKAP - (Production)
UserID :- 00000000 (root)
Program Name :- mq_test
Process :- 00025606
Thread :- 00000001
QueueManager :- TEST
Major Errorcode :xecF_E_UNEXPECTED_SYSTEM_RC
Minor Errorcode :- OK
Probe Type :- MSGAMQ6119
Probe Severity :- 2
Probe Description :- AMQ6119: An internal MQSeries error has occurred
('unknown value of errno' from shmat.)
Arith1 :- 147360 23fa0
Comment1 :- 'unknown value of errno' from shmat.
.
.
.
strerror from shmat
ff1221f0 546F To
ff122200 6F206D61 6E79206F 70656E20 66696C65 o many open file
ff122210 73 s
[ This Message was edited by: eindano on 2001-09-27 01:49 ] |
|
| Back to top |
|
|
| kolban |
| Posted: Thu Sep 27, 2001 6:23 am Post subject: |
|
|
Grand Master
Joined: 22 May 2001
Posts: 1072
Location: Fort Worth, TX, USA
|
| Looking at reports of previous similiar problems, this appears to be caused by the kernel settings being too low. Specifically, the culprit seems to be the SHMSEG attribute. Can you increase this value and see if it makes a difference? Perhaps you could post the kernel settings for your machine back here and we can try and spot other settings that may cause problems? |
|
| Back to top |
|
|
| eindano |
| Posted: Fri Sep 28, 2001 1:08 am Post subject: |
|
|
Newbie
Joined: 26 Sep 2001
Posts: 3
Location: Sweden
|
| We tried to increase the SHMSEG parameter 10 times and our program survived 10 times longer. We found out that the MQCONNX call leaks aprox 1kb per call. Is this a known problem? |
|
| Back to top |
|
|
| kolban |
| Posted: Fri Sep 28, 2001 7:24 am Post subject: |
|
|
Grand Master
Joined: 22 May 2001
Posts: 1072
Location: Fort Worth, TX, USA
|
Are you 100% sure that you are cleaning up your program between MQCONNX calls? Are you certain your are closing queues and disconnecting from the queue manager?
What releases of software (includings PTFS) do you have applied? If you would like to post your program, I'd be happy to take a look-see. If it is too big, you may send me a ZIP file and I will place on my website and add a link to it in the forum ... as always, do NOT send any materials that you would not be willing to share publicly.
You may also wish to open a PMR with IBM defect support if you are sure that you are cleaning up correctly. |
|
| Back to top |
|
|
| eindano |
| Posted: Sun Oct 07, 2001 10:51 pm Post subject: |
|
|
Newbie
Joined: 26 Sep 2001
Posts: 3
Location: Sweden
|
We have talked to IBM support and this is a memory leakage in the MQ libs, a patch will be available as soon as possible.
|
|
| Back to top |
|
|
| bower5932 |
| Posted: Tue Oct 09, 2001 8:21 am Post subject: |
|
|
Jedi Knight
Joined: 27 Aug 2001
Posts: 3023
Location: Dallas, TX, USA
|
| Just for reference, did IBM support give you an APAR number or PMR number to track your problem? If so, could you post it? |
|
| Back to top |
|
|
|
|
