Re: High %wio problem

• What new processes become active around 1340 (when %wio ramps up)?
• Where is IO occuring? Use `sar -d` for this. What resides there? Is "there" a file system or raw partition? If a file system, what files are there and what do they contain? Use fuser to see what processes are involved.
• How are system resources being used during that time? What are the related applications doing? On HP-UX you'd use glance and tusc. I'm not sure what a similar tool would be on Sun.

For example: Heavy writes (expensive on RAID 5)? Ratio of logical to physical IO? System wait states?

Ad infinitum.

Regards,
Eric Stahl

"Dave Weeks" dave_at_tc3.co.uk wrote:

>Hi all,
>
>Quick question regarding high %wio. Every month we get a day where the %wio
>is so high it brings the Oracle server to a crawl so I'm wondering what the
>reasons are that can cause this (eliminated insufficient rollback segments,
>tablespace sizes, and archiving of redo logs). Restarting Oracle cures the
>%wio back to normal.
>
>We have Oracle 8.1.5 EE running on Solaris 7 (5x9Gb disks, raid 5) and have
>a 10Gb database.
>
>Thanks in advance, Dave Weeks.
>
>And here's the output of the 'sar' utility over a typical problem day:
>
>00:00:01 %usr %sys %wio %idle
>01:00:01 0 0 1 99
>02:00:02 0 0 1 99
>03:00:01 1 2 2 98
>04:00:01 0 0 6 94
>05:00:02 0 0 2 98
>06:00:01 0 0 1 98
>07:00:01 0 0 1 98
>08:00:01 0 0 1 98
>08:20:01 0 0 1 99
>08:40:01 0 0 2 97
>09:00:00 0 0 1 98
>09:20:01 0 0 1 99
>09:40:00 1 1 7 90
>10:00:01 0 0 2 98
>10:20:01 0 0 3 97
>10:40:01 1 1 4 94
>11:00:01 0 0 2 98
>11:20:01 1 1 6 92
>11:40:01 1 0 3 96
>12:00:01 0 1 5 93
>12:20:01 0 1 3 96
>12:40:01 0 0 2 98
>13:00:01 0 0 2 98
>13:20:01 0 0 2 98
>13:40:03 0 8 23 68
>14:00:25 0 7 14 79
>14:20:03 0 8 18 75
>14:40:10 1 9 20 70
>15:00:02 0 7 16 76
>15:20:04 0 7 15 78
>15:40:03 0 7 14 79
>16:00:29 0 8 17 75
>16:23:58 1 10 34 54
>16:40:06 2 16 65 18
>
>Average 0 2 9 89
Received on Fri Dec 22 2000 - 00:08:08 CST