Very Computer

We are using Progress 7.3A15 on SCO UNIX 3.2v4.2 on a machine with 96
Mbyte memory.  I am trying to get one of the databases to use 20 Mbytes
or so of shared memory.

I do not seem to be able to set -B (buffers) server start-up parameter
above about 3000 (or about 4 Mbytes), even with SHMMAX set to 8 or 16
Mbytes, without running into problems when trying to start (self
service) clients.  (Progress errors 1147 or 1149).

Progress tech support tell me that there is a SCO recognised SCO bug
with SHMMAX, and that I should set SHMMAX no greater than 4 Mbytes and
increase SHMSEG to compensate.

Can anyone comment (usefully) before I start mucking about?
It will mean another late evening wasted if it does not work.

--
Charles

Turnpike evaluation. For information, see http://www.turnpike.com/

: We are using Progress 7.3A15 on SCO UNIX 3.2v4.2 on a machine with 96
: Mbyte memory.  I am trying to get one of the databases to use 20 Mbytes
: or so of shared memory.

: I do not seem to be able to set -B (buffers) server start-up parameter
: above about 3000 (or about 4 Mbytes), even with SHMMAX set to 8 or 16
: Mbytes, without running into problems when trying to start (self
: service) clients.  (Progress errors 1147 or 1149).

: Progress tech support tell me that there is a SCO recognised SCO bug
: with SHMMAX, and that I should set SHMMAX no greater than 4 Mbytes and
: increase SHMSEG to compensate.

: Can anyone comment (usefully) before I start mucking about?
: It will mean another late evening wasted if it does not work.
:  

Tech support is telling you to use smaller segments, but more of them.  I
can echo the problem on SCO, and suggest following their advice.

HTH.
==========================================================================
Geoff Crawford                                  Phone:    (201) 627 - 0307
Innovative Client Servers                       FAX  :    (201) 627 - 0634

Denville NJ 07834

Thanks to the several who have offered advice and commiseration.

There is indeed a recognised problem with shared memory in SCO 3.2v4.2.
Support Level Supplement UOD397A is available from the SCO BBS and ftp
services to fix it.  I installed it this evening.  It seems to work.

Not having tried it before, I am old enough to find the whole idea of
obtaining a working kernel patch over the net quite impressive.  (But
then I know someone who has successfully uuencoded and e-mailed a
complete Linux kernel).

Setting the Progress server startup parameter -B to 5000 now works,
giving just over 6000 k of shared memory.  However, a value of 10000
(which equates to just under 12 Mbytes in our case) still gives error
1179 when starting a client.

I did not record the text of this error message (too much of a hurry to
get home to drying lasagne) and the Progress on-line help does not know
about it.  The message says words to the effect that the process size is
too big, or something like that.  I suspect that there is another kernel
parameter to tune.

--
Charles

(Sorry about the addition below, it is inserted by the
connection software I am trying, and I have yet to work
out how to get rid of it :(.

Quote:

>Setting the Progress server startup parameter -B to 5000 now works,
>giving just over 6000 k of shared memory.  However, a value of 10000
>(which equates to just under 12 Mbytes in our case) still gives error
>1179 when starting a client.

I should have said *seems* to work.  I arrived at work this morning to
be greeted with several screens containing error messages - including
"forgot lock".  (Brilliant, if you know you forgot it, can't you go back
and get it?)  Under advice from Progress, an attempt was made to rebuild
the index, which showed that the before image was corrupt.  So, I had to
restore the databases from tape, reset -B back down to the pathetic 3000
at which this sorry saga started, and start again.  Any more clues?  
--
Charles

Turnpike evaluation. For information, see http://www.turnpike.com/