HP3000-L Archives

December 1997, Week 3

HP3000-L@RAVEN.UTC.EDU

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Lee Gunter <[log in to unmask]>
Reply To:
Lee Gunter <[log in to unmask]>
Date:
Thu, 18 Dec 1997 16:18:17 -0700
Content-Type:
text/plain
Parts/Attachments:
text/plain (41 lines)
A week or so ago, I posted a summary of a performance problem we were
experiencing following a major installation of Y2K-related changes.  The
Glance/iX monitor was displaying processes experiencing major BLKCB
(control block) waits -- typically displayed as IMP on the global screen --
such that up to 75%-85% of a process's run length was spent in this state.
My obvious first choice for a cause was file or database locking
contention; however, many of the processes shown as concurrently blocked
were not sharing files, and locking sequences or strategies had not
suddenly changed.

Finally, thanks to major assistance from Stan Sieler, the problem was
identified and traced back to a COBOL compiler directive we installed
globally (into COBCNTL.PUB.SYS) a few weeks prior to the onset of the
problem.  This directive -  $control symdebug - was implemented to insert
tracing code necessary to run a new symbolic debugger (TRAX) for future
problem resolution in suspect code.  Unfortunately, the effect was to cause
many processes to become interrupted often (I'm still not sure why ...),
greatly slowing  run times -- primarily for batch processes.  Gratefully,
our online system had not been compiled with this directive, so the effects
could have been much worse, but I'm not sure those processes didn't
occasionally suffer.  Because this directive had been installed several
weeks prior to the Y2K turnover, I didn't even suspect it.

After the problem was initially identified, we began recompiling those
programs exhibiting the worst behavior, and the effect was dramatic.
We've removed '$control ... symdebug' from COBCNTL and from any source with
specific $control directives, and it will only be used as a specific
compile option for programs needing breakpoint debugging -- and then only
on one of the development systems as much as possible.

Many thanks to those who responded -- you were great !

Lee Gunter
Regence Blue Cross Blue Shield of Oregon / Regence HMO Oregon

mailto:[log in to unmask]
voice...503-375-4498    fax.....503-375-4401
==========================================================
The opinions expressed, here, are mine and mine alone, and do not
necessarily reflect those of my employer.

ATOM RSS1 RSS2