Operating System - HP-UX
1748156 Members
3793 Online
108758 Solutions
New Discussion юеВ

Re: Disk utilization on 9i oracle.

 
Charles Temple
Advisor

Re: Disk utilization on 9i oracle.

Sorry, the dbc_max_pct is 10 and min is 5
Yoann Mainguy
Occasional Advisor

Re: Disk utilization on 9i oracle.

You have about the same stats as ue. I'm not a unix or Oracle expert but I don't believe they indicate that the EVA is performing poorly.

I think we tried changing the scsictl queue_depth thing and it didn't change anything.

Simon.
Alzhy
Honored Contributor

Re: Disk utilization on 9i oracle.

Unless I miss it somewhere, you did not mention IF you have performance issues. If you do - it will manifest itself with your queues starting to depart from 0 and yur wait times on your EVA LUN(s) will start to increase as reckone from "sar -d".

The default Glance / (parm?) config file does not do justice to accurately getting real I/O subsystem performance. You will need to tweak it so it matches your storage architcture. Please rely on "sar -d" and "glance -u" to give you a better picture of your I/O load.


Hakuna Matata.
Charles Temple
Advisor

Re: Disk utilization on 9i oracle.

What are acceptable values for sar -d? When would it appear to look like there are issues? We are going live with the boxes and want to try to be ahead of the users calling in if there are performance issues.

Thanks again for all responses.

Charlie
Alzhy
Honored Contributor

Re: Disk utilization on 9i oracle.


avque -- should be no more than 1.0
avwait -- should be 0 or near 0.
avserv -- should be less than 20.

If your sar -d stats meet the above guidelines, then overall - your I/O subsystem should be responsive enough.

Hakuna Matata.
Eric Antunes
Honored Contributor

Re: Disk utilization on 9i oracle.

Nelson (escuse me Charles for using you post),

I've a 5 average value for avwait: can you give me some clue about the reason or anything I should check?

Best Regards,

Eric Antunes
Each and every day is a good day to learn.
Yoann Mainguy
Occasional Advisor

Re: Disk utilization on 9i oracle.

Our stats:

device %busy avque r+w/s blks/s avwait avserv
Av c5t6d0 1.08 0.50 2 12 0.00 5.40
Av c8t6d0 0.96 0.50 2 11 0.00 5.12
Av c19t0d2 80.18 0.50 270 4472 0.00 5.14
Av c19t0d3 82.29 0.50 268 4281 0.00 5.40
Av c19t0d0 0.32 0.50 1 6 0.00 3.79
Av c19t0d1 0.08 0.50 0 1 0.00 2.88

You can see avwait is 0 and the avserv time is the same for all the volumes. All our data/indexs/redo/undo/archive is on c19t0d2 and c19t0d3


Alzhy
Honored Contributor

Re: Disk utilization on 9i oracle.

Eric,

On my EVA connected systems.... avwait is consistently 0. Only if the EVA is swamped to I see it (along with avque) go beyond 0.

What storage subsystem are you using? What other sar -d stats do you observe when you've youe avwait at 5?
Hakuna Matata.
Charles Temple
Advisor

Re: Disk utilization on 9i oracle.

Here is my sar -d 5 5 stats.

2:40:37 device %busy avque r+w/s blks/s avwait avserv
12:40:42 c0t6d0 3.37 0.50 6 46 6.17 14.64
c7t6d0 2.18 0.50 5 42 5.89 12.92
c58t0d6 1.19 0.50 20 354 4.69 0.57
c22t1d2 0.20 0.50 2 25 6.10 1.08
c22t0d4 0.40 0.50 2 29 5.67 0.30
c22t0d7 0.99 0.50 21 370 4.75 0.47
c58t1d1 0.20 0.50 1 19 5.71 0.26
12:40:47 c0t6d0 5.85 0.50 26 219 5.14 3.83
c7t6d0 2.02 0.50 4 36 5.04 9.98
c58t0d6 0.20 0.50 2 35 4.70 0.30
c22t1d2 1.61 0.50 23 371 5.17 0.54
c22t0d4 0.81 0.50 16 276 4.66 0.45
c22t0d7 0.20 0.50 2 29 5.38 0.28
c22t0d6 0.20 0.50 1 11 9.32 0.28
c22t0d5 5.85 0.50 14 287 5.13 4.37
c58t1d1 0.40 0.50 2 32 4.74 1.91
c58t0d0 0.81 0.50 8 187 5.36 1.11
RAC_1
Honored Contributor

Re: Disk utilization on 9i oracle.

For Hein,

I agree to you to some extend, but not completely. while it is true that 100 % i/o is not a accurate measure of EVA(for that matter for SAN), but it sure that those disks
for that LUN are hitting high. what I mean is if SAN lun consists 4 pvs and i/o from Os looks 100 %, it certainly sure that few of them are getting hammered. to further analyze it you can use iostat on individual disks, you can use array commands to check thode disks. The point is, it would be possible to pinpoint the disk that is hitting high and then see what files are being present on it and carry it forward.

But I do agree that statspack and i/o stats will go a long way to help on this issue.

Anil
There is no substitute to HARDWORK