T.R | Title | User | Personal Name | Date | Lines |
---|
687.1 | Check REDO log disk | MAIL::DUNCANG | Oracle... the one-line database | Fri Jul 13 1990 16:33 | 12 |
| VMS does what Oracle tells it to so the problem is probably not
related to VMS so don't let the Oracle guys come back with the "it's
a VMS problem" crap.
I assume you're talking about Oracle V6. With that in mind, when
Oracle V6 commits, there is ALWAYS i/o to the disk where the redo logs
are located. (The actual data pages are written to disk later.)
If you are experiencing excessive DIO, check the size of the redo logs
and the various Oracle parameters that relate to log checkpoint
intervals.
-- gerry
|
687.2 | Yes, but why? | NZOV07::HOWARD | NZ: Where Digital's Week Begins | Tue Jul 17 1990 01:50 | 18 |
| Well the O people have admitted that their software calls the VMS
system service used to create temorary files, and is passed null
parameters hence the funny file names.
The local office is still not sure WHY this is done and are "following
it up".
Interestingly we have increased user working set extents (WSEXTENT)
to 20,000 so it is not for lack of memory. Peak used so far is
9950 (thats 5MB !). Physical memory is only 60% used, as we are allowing
for more users.
A background process named SRV has been running with 2048 WSEXTENT.
The O people have only just found the place in the ORACLE manual that
explains how to increase that, so we wait to see if things get any
better.
Cheers, Martin
|
687.3 | Better? | TRCA01::MCMULLEN | | Tue Jul 31 1990 21:57 | 5 |
| Martin,
Did they get any better?
Ken M
|
687.4 | | NZOV07::HOWARD | NZ: Where Digital's Week Begins | Fri Aug 03 1990 14:24 | 30 |
| � Did they get any better?
Not yet.
The O people have said that the problem will be present at ALL VMS
sites (and possibly those on other platforms). They are asking
the head office to look into ways of either providing a workaround
to bypass the hard-coded logic, or to provide a modification.
No progress at the customer site so far.
It seems that my customer processes large batches of invoices which
cause the working set for the processes to just keep increasing.
If they had smaller batches, and exited more often, the problem
would not be seen.
The local O people (when prompted by the customer) admitted that
other customers may have the problem and not realise it.
SO IF YOUR CUSTOMER IS HAVING PERFORMANCE PROBLEMS WITH ORACLE
FINANCIALS, do a SHOW DEV/FIL and look for those funny files.
Also use SHOW PROC/CONT to monitor the process details.
As an aside; it's interesting to see the positions at the customer
meeting. Some at Digital are very tolerant of customer demands
through fear of losing the box sale. Oracle are very slack in support
letting issues drag on for weeks, but don't have to worry to the same
degree becuase for them the customer is captive.
Cheers, Martin
|