index search exceeds record limit of (100000) abend

Elaine Pinnington · New User Joined: 07 Dec 2007 Posts: 6 Location: France

Hi,

I have a CICS/VSAM program that searches a file using various indexes based on criteria supplied by a user. The problem is that every so often the program abends because the limit of 10000 records searched has been exceeded. Is there a way of intercepting this count or abend so that we send an error message back to the front end and log the search criteria that they were using???
We have no control over the front end and the search criteria are not very well validated.

It is a very long time since I have worked with CICS and VSAM (abt 15 years).

Thanks

Bill O'Boyle · Posted: Fri Sep 21, 2012 8:34 pm

Suspicious. Are you using a subscript, defined as S9(04) COMP and the program was compiled using the TRUNC(OPT) compiler option?

If true, define the S9(04) COMP field as S9(08) COMP and you'll be OK.

If your compiler support COMP-5 (Native Binary) you can keep the S9(04) variable, just redefine it as COMP-5 or just redefine it as S9(08) COMP-5, which is even better.

Indices (as opposed to subscripts) do have some type of limit (probably X'7FFFFFFF'), but certainly not 100000....

Bill O'Boyle · Posted: Fri Sep 21, 2012 8:46 pm

Elaine,

Your post-header indicates 100000 (One Hundred Thousand) but your post-text indicates 10000 (Ten Thousand).

Typo?

Bill Woodger · Posted: Fri Sep 21, 2012 8:54 pm

Isn't this going to be some "limit" set in a program somewhere?

dbzTHEdinosauer · Posted: Fri Sep 21, 2012 8:55 pm

apologize for raining on your parade Elaine,
but

Bill O'Boyle · Posted: Sat Sep 22, 2012 2:58 am

Elaine,

You can dynamically calculate the maximum OCCURS without knowing it, provided that the table be structured similar to the following -

Bill Woodger · Posted: Sat Sep 22, 2012 4:05 am

Just out of interest, this works as well:

Bill O'Boyle · Posted: Sat Sep 22, 2012 4:21 am

Bill,

I began using this calculation with the introduction of COBOL II, in the mid-80's. Old habits die hard.

dick scherrer · Posted: Sat Sep 22, 2012 9:03 am

Hello,

Elaine Pinnington · New User Joined: 07 Dec 2007 Posts: 6 Location: France

Hello,

My appologies for the poor explanantion.

The program is a CICS module that looks up bank transaction detail from a VSAM file. In certain cases it does a readprev on an alternate index on said VSAM file. Under certain circumstances (I assume the readprev has searched through more than the threashold of 100000 records and still not found what it is looking for) the CICS module abends with a threashold limit exceeded.
My French collegues have tried to narrow down the search criteria so that this abend doesn't happen but it still happens fairly often.
I figured that it would be better to trap the abend and handle it in a controlled manner rather than fiddle with the search criteria and then analyze the search that are causing the problem.
At the moment the program just checks EIBRESP for NOT FOUND and EOF. Noone here has any CICS/VSAM experience. I tried googling for any info on this abend but with no luck.
I figure that if all else fails I'll just trap add a test for any other error (other than NOT FOUND or EOF) and handle it as a threshold exceeded error.
the code sample is this ...

dbzTHEdinosauer · Posted: Tue Sep 25, 2012 7:16 pm

why not start with a START BROWSE, then do a READ PREV?

and who/what is generating the 10000(0) index search error message?????????
don't think that it is the op-sys, i think it is a user written error.

Robert Sample · Posted: Tue Sep 25, 2012 7:55 pm

dbzTHEdinosauer · Posted: Tue Sep 25, 2012 8:20 pm

of the possible conditions:
DUPKEY,
ENDFILE,
FILENOTFOUND,
ILLOGIC, |
INVREQ,
IOERR,
ISCINVREQ,
LENGERR,
LOCKED,
NOTAUTH,
NOTFND,
RECORDBUSY,
SYSIDERR

the code checks for ENDFILE and NOTFND.
does the code following suggest that the resp is ok
if it is not the two mentioned?

also, what is this stuff about walking thru an index?
readprev presents the record or not.
what search is going on??????

Elaine Pinnington · New User Joined: 07 Dec 2007 Posts: 6 Location: France

They do start with a start browse, then a read prev.

The program abends like this .....

Abend RLFC in transaction FRFX Pgm(TSFTVWH1) occured on CICS CIBDFPA9

Threshold exceeded

File control requests exceeded. > (100000 request VSAM)

dick scherrer · Posted: Tue Sep 25, 2012 8:48 pm

Hello,

Can you count how many entries there are for the "key"(s) that run amok?

Possibly the program has gone into a loop and is re-reading things over and over - until the threshold is exceeded?

dbzTHEdinosauer · Posted: Tue Sep 25, 2012 9:12 pm

of course the program builds some kind of key
and then also builds a low-key as a stop?

and DFHRESP(NORMAL) should be the only test after the READPREV.
anything else should be treated as an exception.

Elaine Pinnington · New User Joined: 07 Dec 2007 Posts: 6 Location: France

Hello,

I thought that when you pass a key to a read prev it would go off and read through the index until it finds a match (or not). The problem here is that with certain keys it is possible that the keyed read goes past the threshold of 100000 records (causing an abend) before it finds a match or reaches EOF. I was hoping that there would be some way of trapping this abend and handling it.

If I just add a catch-all that picks up on any EIBRESP that comes back other than OK, EOF and NOT FOUND then I at least capture some more information and handle the error in a more controlled manner.

With an error like this, does it return to the program with some kind of error code or does CICS just end the program ???

We can't recreate this error in DEV or UAT so can't just experiment!!!

Thanks

Robert Sample · Posted: Wed Sep 26, 2012 4:29 pm

dbzTHEdinosauer · Posted: Wed Sep 26, 2012 4:39 pm

your program (in a loop) is issuing (executing) 100000 readprev.

upon a successful READPREV,
the RIDFLD is updated with the "id" of the current record.

apparently some compare is being made to the expected id
and the returned id.

now, we go back to a post I made earlier,
when the starting "key" is built, an ending key should also be built
otherwise you will never find an equal to the starting key
and keep reading until eof (start of file in this case).

you should also include a compare with the built key for out-of-range.

at some point you should realize that what you are looking for is not there,
and to continue reading will never result in a FOUND condition.

or you could simply implement a counter in your program
and at 99999 send a message back to the user,
not found and thus not issue the 100000th READPREV that would result in an abend.