The support folks have information about what should be where in memory. So its reasonable that they would request core dumps for analysis. As I noted earlier, memory usage for/by DataStage processes is not public knowledge (and tends to change release by release).
> ----------
> From:
muthusamy.karthikeyan@db.com[SMTP:
muthusamy.karthikeyan@db.com]
> Reply To:
informix-datastage@oliver.com
> Sent: Wednesday, 15 November 2000 19:46
> To:
informix-datastage@oliver.com
> Subject: RE: debugging core files
>
> Ray,
>
> > Use any core inspection tool you have available. However, without
> > information about memory usage in DataStage - which is not published
> > and will not be - you have no idea what youre looking at, so
> > attempting to analyse core files is somewhat moot. If the job still
> > has a status of Running, you may find files in the &PH& subdirectory
> > containing the name StageRun. These are the processes that execute
> > the code produced by Transformer stages. There will also be a
> file
> > for the job itself, with DSD.RUN as part of its file name. The
> > object for the job control and Transformer stage executables is
> found in
> > RT_BPnn.O, where nn is the job number. This is compiled BASIC - not
> > operating system executable code. What were you planning to do with
> > it?
>
> We were facing the mentioned problem quite frequently. And, so i was
> curious to find whether i could get some info on whether there is any
> problem in respect with AIX ( memory, permissions..et) or with the
> data stage. when we escalated the problem to support, they were asking
> for the core file. so i thouhght i could try my hand in breaking the
> core file.
>
> david,
>
> >> Is this hanging predictable or repeatable, ie does it always happen
> when you
> >> run a particular job?
> >> Or, is it random, perhaps more likely to happen when there are a
> >> number
> of
> >> jobs running? Does this job have a large number of stages?
>
> The haning is not predictable , it happens randomly on different jobs.
> To some extent, yes, most probably hangs occur when a number of jobs
> are running. and, at the max. we have around 10 stages. since we run a
> large number of jobs in production, we are not able to keep track of
> jobs which had been hanging for a logn period of time, which in turns
> affects the up time.
>
> any way out to crack this hang problem, or the cause of it?. any way
> to get in to universe to check the dead ones ?. does the data stage
> kill the jobs that had hanged ( but not updating the status, which
> shows running), bcoz we could not find any process running for the
> hanged one.
>
> thankx for your attention.
>
> Thanks and Regards
> Karthik
> Analyst
> Deutsche Bank
> singapore
> 065-423-7410
>
>