[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference orarep::nomahs::rdb_60

Title:Oracle Rdb - Still a strategic database for DEC on Alpha AXP!
Notice:RDB_60 is archived, please use RDB_70..
Moderator:NOVA::SMITHISON
Created:Fri Mar 18 1994
Last Modified:Fri May 30 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:5118
Total number of notes:28246

5058.0. "Number of buffers used" by svrav1.au.oracle.com::MBRADLEY (I was dropped on my head as a baby. What's your excuse?) Fri Feb 21 1997 00:07

I was out on a site today reviewing a monitor log file where users had been 
recieving error on attach because there were insuficient number of global 
buffers available.

The funny thing is that the database only allows 100 users, the number of 
global buffers is 3500, and the user limit is 30.

This appears to happen when many (around 20) recovery processes are 
started.

I can understand why they would fail the attach for the 101'st user, but 
how could they possibly use more than 3000 buffers (user limit*number users 
which is 30*100) with less than 100 attaches?

G'day,

Mark.
T.RTitleUserPersonal
Name		DateLines
5058.1M5::JHAYTERFri Feb 21 1997 10:227
>This appears to happen when many (around 20) recovery processes are 
>started.

what is the "Default recovery buffer count"?
I don't know if recovery processes can exceed the user limit.
5058.2NOVA::R_ANDERSONOracle Corporation (603) 881-1935Fri Feb 21 1997 15:425
Well, the obvious guess would be that DBR requires buffers in addition to the
dead user's (who isn't really quite dead yet :-), and DBR requires MORE buffers
(typically) than the dead user.

Rick
5058.3It didn't appear sosvrav1.au.oracle.com::MBRADLEYI was dropped on my head as a baby. What's your excuse?Wed Feb 26 1997 21:1612
>Well, the obvious guess would be that DBR requires buffers in addition to the
>dead user's (who isn't really quite dead yet :-), and DBR requires MORE buffers
>(typically) than the dead user.

This does not appear to be the case. The monitor log shows the failed user 
relinqishing his buffers prior to the DBR grabbing them. I will see if I 
can get an extract of the monitor log file to show you, but that may take a 
few days.

G'day,

Mark.
5058.4Just taking aother look at thissvrav1.au.oracle.com::MBRADLEYI was dropped on my head as a baby. What's your excuse?Sun Mar 16 1997 21:0722
Rick,

>   <<< Note 5058.2 by NOVA::R_ANDERSON "Oracle Corporation (603) 881-1935" >>>
>
>Well, the obvious guess would be that DBR requires buffers in addition to the
>dead user's (who isn't really quite dead yet :-), and DBR requires MORE buffers
>(typically) than the dead user.

Well in my testing, the DBR always uses the same number of buffers as the 
failed user. The only thing that would make sense is if the failed user was 
on another node. Then you would get the DBR taking additional buffers on 
the recovery node.

But I still can't see how we could run out of buffers when the limit seems 
to be the number of users.

I am getting them to look at the monitor log to see if the failed user was 
on the same node as the DBR, but either way it doesn't seem to make sense.

G'day,

Mark.