Re: How to improve oracle IO performance ?

On Feb 25, 6:14 am, "jshen...._at_gmail.com" <jshen...._at_gmail.com> wrote:
> hi,
>
> Our oracle8.1.7 runs on HP-UX 11i with HP disk-array xp-1024. The
> disk array is raid-0.
>
> there are about 3 million accout mainteined in the DB which has
> accouting data input into the db. There are three snapshots on four
> tables of the DB. The refreshing interval is 5min. the four tables
> which are snapshoted mainteins the most part of information in DB.
>
> The max session number of DB is 600, and the peak connection number
> recorded is 571.
>
> we experience DB performance problem in past months. And, there
> happened three times of DB crash.
>
> I'm new to this area. Is there anyone would like to share
> knowledge and exprience with such problem ? ( e.g. where should I
> start ? what should be the key point of analysis? is there any must-
> read on DB optimization? )
>
> any word will be highly appreciated.
>
> Joe
>
> ps. the following is tracefile of DB crash , and disk io on DB server.
>
> 1. DB crash trace file
>
> /oracle/app/oracle/admin/pin65db/bdump/pmon_8454_pin65db.trc
> Oracle8i Enterprise Edition Release 8.1.7.4.0 - 64bit Production
> With the Partitioning option
> JServer Release 8.1.7.4.0 - 64bit Production
> ORACLE_HOME = /oracle/app/oracle/product/8.1.7
> System name: HP-UX
> Node name: cm3-dcb2
> Release: B.11.11
> Version: U
> Machine: 9000/800
> Instance name: pin65db
> Redo thread mounted by this instance: 1
> Oracle process number: 2
> Unix process pid: 8454, image: oracle_at_cm3-dcb2 (PMON)
>
> *** 2007-02-08 11:22:51.273
> *** SESSION ID:(1.1) 2007-02-08 11:22:51.273
> Exception signal: 10, code = 800003ffff7f3d88
> sc_onstack: 00000000172bc000, sc_mask 0000000000000000
> ORA-07445: exception encountered: core dump [10] [4286528904]

In addition to what Mark said, look around on metalink for articles about ORA-7445,
especially Notes 153788.1 211909.1 314422.1

Mark's probably right, usually a core dump means some OS limit has been exceeded. You can figure out what the 10 is by finding a file named signal.h and looking for a value of 10 defined. It can also mean your hardware is starting to have problems - do you have monitoring of them (on mine alerts get mailed to root, except events that are confined to the arrays)? It can also mean you have patch issues, for example, you need some HP patch for your array. A perusal of the HP patches can be frightening, though usually the "latest patches" seem to have all of them addressed once you work your way through them.

Also, is there anything besides normal log switching in the Oracle alert file?

jg

--
@home.com is bogus.
"Get Less" http://www.gearlog.com/2007/01/tmobile_disses_opera_says_get.php