I beg to differ! A commit takes place under the scope of a connection handle. When a channel exit connects to the queue manager, it will get back an HCONN, but the reason code will be MQRC_ALREADY_CONNECTED and the HCONN it gets is the same one the MCA is using, so any commits issues by the exit *will* interfere with the commits issued by the MCA.
David -----Original Message----- From: MQSeries List [mailto:[EMAIL PROTECTED] Behalf Of Ruzi R Sent: 03 September 2003 12:15 To: [EMAIL PROTECTED] Subject: Re: MVS msg exit (MQPUT) processing causes increase in MSTR region size. > be careful with the suggestion from Ruzi. > > The message exit runs inside the unit of work for > the MCA. If you commit, > then you are committing the channel's work as well. Not true. The application's commit is separate from Channel's commit. Channel's commit depends on BATCHSZ/BATCHINT attributes. Ruzi --- Neil Casey <[EMAIL PROTECTED]> wrote: > Hi Dax (are you a DS9 fan?) > > be careful with the suggestion from Ruzi. > > The message exit runs inside the unit of work for > the MCA. If you commit, > then you are committing the channel's work as well. > > You shouldn't have to worry about not committing in > your exit as the MCA > will issue a commit for every batch. > > I have no idea why your real storage usage is rising > only when your exit is > run, but to test it out, try running your programs > without any exits, and > at the same time run a local program which inserts > the same message load as > the exit would do. Look at the memory usage after > this test and compare > with running your exit to determine if your exit is > behaving badly. I would > also suggest checking the virtual usage as well as > real. If the virtual > usage is unchanged, then what you are seeing is not > a leek, but a change in > the working set in an unconstrained system. This > would represent normal > behaviour. Your system is clearly unconstrained > because your paging rate is > 0. > > Regards, > > Neil Casey. > > > |---------+----------------------------> > | | Ruzi R | > | | <[EMAIL PROTECTED]| > | | M> | > | | Sent by: MQSeries| > | | List | > | | <[EMAIL PROTECTED]| > | | n.AC.AT> | > | | | > | | | > | | 03/09/2003 02:29 | > | | Please respond to| > | | MQSeries List | > | | | > |---------+----------------------------> > > >--------------------------------------------------------------------------- -----------------------------------| > | > > | > | To: [EMAIL PROTECTED] > > | > | cc: > > | > | Subject: Re: MVS msg exit (MQPUT) > processing causes increase in MSTR > region | > | size. > > | > > >--------------------------------------------------------------------------- -----------------------------------| > > > > > The first thing that came to my mind (without > reading > all the deatils carefully-- not that I could > understand the code all that much) is that, it might > be caused by a delayed "commit"??? From your code, I > think you are using the default MQPMO. On the > mainframe the Syncpoint is the default ( > Syncpoint=yes). Looks like you are not commiting > each > message after each PUT (or after so many messages). > When you diconnect without the explicit commit, an > implicit commit occurs committing all the 1000 > messages. In the meantime, the 1000 messages > (depending on on their size probably) may be causing > the space to expand. Since you are experimenting, I > suggest you try using a commit after every (or say > 10 > messages) and see whether you get the some problem. > > I don"t know if the above is causing your problem, > but > I thought I would share my opinion with you anyway. > > Best regards, > > Ruzi > > --- d a <[EMAIL PROTECTED]> wrote: > > To the MVS people on the list.... > > > > I have been playing around with the following > > message exit code it all > > works fine BUT, I have just noticed that the MSTR > > address spaces for the 2 > > queue mgrs I am testing on expands considerably > not > > a good thing !!!! > > > > Here is the setup, I have 2 queue mgrs QMG1 and > > QMG2, and a set of SDR/RCVR > > channels between them, all with the exit defined > as > > a message exit on them. > > I then send through 1000 messages in both > directions > > to drive the exit > > defined on the RCVR and SDR channels. > > > > The result Note the Real storage frames: > > > > Before processing: > > ----------------- > > JOBNAME ,StepName,ProcStep,JobID ,Owner > > ,C,Pos,DP,Real,Paging, SIO > > QMG1MSTR QMG1MSTR PROCSTEP STC05201 QMG1MSTR NS > > FE 8084 0.00 0.00 > > QMG1CHIN QMG1CHIN PROCSTEP STC05203 QMG1CHIN IN > > FB 2897 0.00 0.00 > > > > JOBNAME ,StepName,ProcStep,JobID ,Owner > > ,C,Pos,DP,Real,Paging, SIO > > QMG2MSTR QMG2MSTR PROCSTEP STC05202 QMG2MSTR NS > > FE 8056 0.00 0.00 > > QMG2CHIN QMG2CHIN PROCSTEP STC05204 QMG2CHIN IN > > FB 2970 0.00 0.00 > > > > > > After processing: > > ----------------- > > JOBNAME ,StepName,ProcStep,JobID ,Owner > > ,C,Pos,DP,Real,Paging, SIO > > QMG1MSTR QMG1MSTR PROCSTEP STC05201 QMG1MSTR NS > > FE 10T 0.00 0.00 > > QMG1CHIN QMG1CHIN PROCSTEP STC05203 QMG1CHIN IN > > FB 3307 0.00 0.00 > > > > JOBNAME ,StepName,ProcStep,JobID ,Owner > > ,C,Pos,DP,Real,Paging, SIO > > QMG2MSTR QMG2MSTR PROCSTEP STC05202 QMG2MSTR NS > > FE 10T 0.00 0.00 > > QMG2CHIN QMG2CHIN PROCSTEP STC05204 QMG2CHIN IN > > FB 3299 0.00 0.00 > > > > > > I have determined that the problem is with the > MQPUT > > processing if I > > comment it out, then the region size of the MSTR > is > > fine. But why it > > happening I don t know. > > > > A few other things to mention: > > > > - The TESTEXIT is non-APF in a non-APF > > authorized library (as it should be). > > - It has been linked with CSQXSTUB. > > - I have validated that the Getmain / > Freemain > > processing I do is okay. > > - Queue mgrs are both 5.3 > > - I replaced the MQOPEN, MQPUT, MQCLOSE with > > an MQPUT1 and got the exact > > same results. > > > > SO why why why is this happening please help is > > you can. > > > > Thanks very much in advance for any assistance. > > > > Thanks Dax > > > > > > The code follows: As mentioned, I have been > playing > > with this so there is no > > real functionality to what follows, other than it > > will write a message to > > queue named: TEST.EXIT.QUEUE each time it is > > invoked, for INIT, MSG, XMIT, > > TERM processing, etc. > > > > > > > > TESTEXIT TITLE 'TESTEXIT - Test exit' > > * > === message truncated === Instructions for managing your mailing list subscription are provided in the Listserv General Users Guide available at http://www.lsoft.com Archive: http://vm.akh-wien.ac.at/MQSeries.archive Instructions for managing your mailing list subscription are provided in the Listserv General Users Guide available at http://www.lsoft.com Archive: http://vm.akh-wien.ac.at/MQSeries.archive
