I guess its bug 1397603 on T64 unix. The workaround is to set

_db_handles_cached = 0

Babu
----- Original Message -----
From: "Mike Ault" <[Email Address Removed] "LazyDBA.com Discussion" <[Email Address Removed] Friday, December 21, 2001 9:50 AM
Subject: RE: buffer busy wait & latch free ( cache buffer chain )


Also:

Generic Bug on 8.1.6-8.1.7 ALL PLATFORMS

During investigation of performance problems here at a client we discovered
an interesting set of stats, the total Shared pool used by code (SQL and
PL/SQL) was only about 30 megabytes, however, the shared pool was showing
100% utilization. Further investigation revealed that the "state objects" in
the shared pool were consuming all available space (in this case a whopping
110 megabytes.) "State objects" are created when AND-EQUALs are used,
especially in Pro-C.

This results in ORA-04031 and can also result in high cache buffer chain
contention.

This is bug 1640583 on Solaris but it is marked as a generic bug on all
platforms and since we are on Tru64 5.0 here, you can see that it probably
is. This effects 8.1.6.0 and will require upgrades or patches on almost all
releases upto 8.1.7.2 where it is fixed. There are also patches for
8.1.6.3.1 It is also fixed in 9i.

-----Original Message-----
From: Jane Mikityanski [mailto:Jane.[Email Address Removed] Tuesday, December 04, 2001 3:16 PM
To: 'Mike Ault'; LazyDBA.com Discussion
Subject: RE: buffer busy wait & latch free ( cache buffer chain )


Mike,
Can you please comment more on this? I guess _db_block_hash_buckets and
_db_block_hash_latches are undocumented parameters.
I have not set those up yet. How risky is it to change undocumented
parameters and where can I read about them?
I have the same problem: there is many "latch free" #66 (which is cache
buffer chain) waits on a system.
I'll also appreciate any tips on how to tune these waits. I've read several
papers, but still no luck in tuning.

Thanks,
Jane Mikityansky
Database Administrator
(781)505-3324


>-----Original Message-----
>From: Mike Ault [mailto:[Email Address Removed] Tuesday, December 04, 2001 7:39 AM
>To: LazyDBA.com Discussion
>Subject: RE: buffer busy wait & latch free ( cache buffer chain )
>
>
>Why would you set freelists this way? usually it is set to the
>umber of transactions you expect to update or inert into a
>single block at the same time. Look at increasing the
>db_block_lru_latches. Make sure _db_block_hash_buckets is set
>to twice the value of db_block_buffers rounded up to the
>nearest prime number and set _db_block_hash_latches to 32k (up
>from its default of 1024).
>
>Mike
>
>-----Original Message-----
>From: VIVEK_SHARMA [mailto:[Email Address Removed] Tuesday, December 04, 2001 4:54 AM
>To: LazyDBA.com Discussion
>Subject: buffer busy wait & latch free ( cache buffer chain )
>
>
>
>In report.txt we are getting the wait
>"buffer busy wait" _& wait on "latch free" for "cache buffer chain"
>
>Could an excessively Large db_block_buffer cause this ?
>
>NOTE - Transaction OLTP in nature
>
>NOTE - freelists is set = 2* cpu_count
>
>Any oher advice ?
>
>
>
>--------
>Oracle documentation is here:
http://tahiti.oracle.com/pls/tahiti/tahiti.homepage
To unsubscribe: send a blank email to oracledba-[Email Address Removed] subscribe: send a blank email to oracledba-[Email Address Removed] the list archive: http://www.LAZYDBA.com/odbareadmail.pl
Tell yer mates about http://www.farAwayJobs.com
By using this list you agree to these
terms:http://www.lazydba.com/legal.html


--------
Oracle documentation is here:
http://tahiti.oracle.com/pls/tahiti/tahiti.homepage
To unsubscribe: send a blank email to oracledba-[Email Address Removed] subscribe: send a blank email to oracledba-[Email Address Removed] the list archive: http://www.LAZYDBA.com/odbareadmail.pl
Tell yer mates about http://www.farAwayJobs.com
By using this list you agree to these
terms:http://www.lazydba.com/legal.html


Oracle LazyDBA home page