Cache Overflow in EB.PHANTOM routine

[@di]

Hi,

I have a certain routine which triggers and runs from EB.PHANTOM i.e.,
it is called from the run routine of the phantom. This routine is
quite a large one and it in turn calls several other local and core
routines. I have been facing frequent cache overflows while running
this.
I tried flagging the routine with .NTX and also increased the cache
limit to 999 but neither had any effect.
Could some one help me with how to solve this problem?

My environment details:
OS : IBM AIX 5.3
DB : jBASE 4.1.5.18
T24 release: R06.007

[Mark Hogden]

You will likely have many more problems until you learn how to read.

IMPORTANT: Type [T24] at the start of the subject line for questions
specific to Globus/T24

[@di]

Hi,

I assumed this would be more of something to do with jBASE transaction
management than T24 and that some jBASE experts on the forum would be
able to help me. This was why I didn't prefix my subject line with
[T24]. Sorry if I was mistaken. I'll watch out next time.

> T24 release: R06.007- Hide quoted text -
>
> - Show quoted text -

[avinash d]

Hi,

 I reckon that you must have called F.READ in the routine. If thats the case replacing it with the READ statement should fix your cache overflows. Do let me know if this fixes your problem.

Nash

[Luis Lascano]

Probably your local routines are doing a lot of F.WRITE and need commit
points. If you do not put these commit points then you have a cache
overflow.

On T24 commit point is the call to JOURNAL.UPDATE

Of course, doing just WRITE then you don't have the problem but could have
data integrity problems in case it fails down in the middle of processing
one message.

Luis

-----Original Message-----
From: jB...@googlegroups.com [mailto:jB...@googlegroups.com] On Behalf Of
@di
Sent: Monday, 12 May, 2008 10:01 AM
To: jBASE
Subject: Cache Overflow in EB.PHANTOM routine

[Richard W]

If the transaction managment is the problem then change the eb.phantom
in to a service.
We also had (still have) experience with memory problems. Mainly to do
when using idesc.

[@di]

Hi,

Changing it into a service seems to solve my problem.

Luis & Avinash,

I'm working on using JOURNAL.UPDATE for committing F.WRITEs to the
database but I reckon it will take a bit of time as it is quite a
large routine.

Could somebody explain this to me :- I've flagged my routine with
a .NTX which in my understanding means that there will be no
transaction boundaries set on this routine and WRITEs will be
committed to the database as and when they occur in the routine. How
is it then, that my routine does not respond to this flag and
continues to fill the buffer? Is it because it is running as a
phantom? Will the fuctionality of a .NTX be realised only when running
a routine under a batch?

> > - Show quoted text -- Hide quoted text -

[Greesh Dhir]

From my understanding the .NTX does perform the writes immediately and hence JOURNAL.UPDATE does not need to be called.  However the Journal common variable record may still being added too and I think this is what's killing the cache.  A call to JOURNAL.UPDATE re-initialises the common journal variable to blank/NULL.
 


 
Regards
 
Greesh


> Date: Thu, 15 May 2008 22:13:45 -0700
> Subject: Re: T24: Cache Overflow in EB.PHANTOM routine
> From: adi...@gmail.com
> To: jB...@googlegroups.com