Re: "WARNING: aiowait timed out 1 times" message and slow DB.What check

Hello!

Which Oracle version are you on?
Which value does your OPTIMIZER_MODE initialization parameter have? There has been some problems with it when it's FIRST_ROWS, especially when doing full import.

Also, you might want to increase log_buffers parameter and log file sizes to get small performance improvement (as seen from v$system_event)

Is the imp process running on a remote server from Oracle instance?

Tanel.

"Kantix" <francesco dot cantisano_at_tin.it> wrote in message news:b2vf8i$1gfnj5$1_at_ID-74096.news.dfncis.de...
> Hi to all,
> I'm experiencing a very slow import (24hours for about 45G) in a brand new
> cluster (sun-cluster 3, Solaris 8), 2 Sun-Fire-480R with 2CPU 900Mz.
> We have all datafiles on a raid 0+1 Veritas filesystem made over 8 disks
> (4+4) seagate 36G 10000rpm.
> The main wait events are (from V$System_Event)
>
> Wait Event Waits Timeouts Time Waited Average Wait
> rdbms ipc message 2144871 925069 246679614 115
> PX Idle Wait 396350 396354 79335267 200
> db file parallel write 15563 0 48594570 3122
> pmon timer 149757 134639 39659564 265
> smon timer 1940 1266 37142397 19146
> SQL*Net mesg from cli 1799654 0 7562073 4
> log buffer space 85072 55741 6774074 80
> direct path write 53854 0 3698614 69
> log file swtch (ckp inc.)34631 34378 3469422 100
>
> I've read the notes abou the WARNING regarding the asyncronous io
> and I'm afraid that something is configured in a not good way;
> what should the sysadmin investigate ad the OS level to confirm that?
> What type of checks we can do?
> The import parameter are:
> buffer=400000000 feedback=2000 commit=N
> I've set the sort_area_size to 128M, the TEMP is locally managed and we
are
> using the UNDOTBS feature,the sga is 2653131224 with 384M of shared_pool.
> The only process running is the import.
> Thanks to all,
> Francesco
>
>
Received on Wed Feb 19 2003 - 10:51:06 CST