Re: iostat output
- From: Matthew Zito <mzito@xxxxxxxxxxx>
- To: oracle-l@xxxxxxxxxxxxx
- Date: Wed, 28 Apr 2004 21:01:56 -0400
Well, unless I mistake its shape and making quite, it is that shrewd
and knavish sprite of an iSCSI implementation on Linux. There's a
couple of potential problems here, so I'll go through some diagnostic
stuff step by step:
-Are you using a hardware iSCSI card? Our internal testing here at
gridapp shows that a dual-cpu host can push roughly (very roughly!) 40
MB/sec. of I/O over a dedicated gige link without hardware offload.
-What is the CPU utilization like on your host side while this I/O is
happening?
-How large are the volumes on the filer side that the luns are being
presented off of? sysconfig -r will show this
-Also, look at the disk utilization on the filer side. use the priv
set advanced command, followed by statit -b (wait a minute or so during
a period of heavy i/o) then run statit -e. Take a look at the data you
get back from the filer?
-Jumbo frames? That should be a requirement.
-Also, iostat on linux is known to lie, especially with nonstandard
block devices - are you actually seeing performance problems? The
iostat does show a _lot_ of I/O.
-Finally, why iSCSI? Why not just use nfs?
-What type of filer?
You can respond to me directly if you like - unless the list at large
is interested in iSCSI on netapp.
Thanks,
Matt
--
Matthew Zito
GridApp Systems
Email: mzito@xxxxxxxxxxx
Cell: 646-220-3551
Phone: 212-358-8211 x 359
http://www.gridapp.com
On Apr 28, 2004, at 8:10 PM, Sai Selvaganesan wrote:
> hi
>
> the following is the o/p of iostat on a new configured machine with
> netapp filer over icsci running oracle 9i database(1 tb in size).
>
> Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s avgrq-sz avgqu-sz
> await svctm %util
> sda 807.31 2.99 252.82 1.33 8486.38 34.55 33.53 8.50
> 33.07 5.23 13.29
> sda1 807.31 2.99 252.82 1.33 8486.38 34.55 33.53 8.50
> 33.07 5.23 13.29
> sdb 2791.03 0.33 1055.81 3.99 30790.70 34.55 29.09 44.12
> 41.13 8.03 85.05
> sdb1 2791.03 0.33 1055.81 3.99 30790.70 34.55 29.09 44.12
> 41.13 8.03 85.05
> sdc 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
> 0.00 0.00 0.00
> sda 6446.67 3.00 1444.00 1.33 63122.67 34.67 43.70 91.73
> 63.49 6.90 99.67
> sda1 6446.67 3.00 1444.00 1.33 63122.67 34.67 43.70 91.73
> 63.49 6.90 99.67
> sdb 2137.67 0.33 587.33 2.33 21826.67 21.33 37.05 32.13
> 54.78 13.23 78.00
> sdb1 2137.67 0.33 587.33 2.33 21826.67 21.33 37.05 32.13
> 54.78 13.23 78.00
> sdc 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
> 0.00 0.00 0.00
> sda 682.67 2.00 201.67 1.33 7080.00 34.67 35.05 14.90
> 73.23 14.61 29.67
> sda1 682.67 2.00 201.67 1.33 7080.00 34.67 35.05 14.90
> 73.23 14.61 29.67
> sdb 11828.67 0.00 2243.33 1.67 112573.33 13.33 50.15 125.23
> 55.68 4.45
> 100.00
> sdb1 11828.67 0.00 2243.33 1.67 112573.33 13.33 50.15 125.23
> 55.68 4.45 100.00
> sdc 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
> 0.00 0.00 0.00
> sda 1878.33 3.00 480.33 1.00 18864.00 32.00 39.26 36.10
> 75.07 20.43 98.33
> sda1 1878.33 3.00 480.33 1.00 18864.00 32.00 39.26 36.10
> 75.07 20.43 98.33
> sdb 11763.33 0.00 1908.00 1.00 109365.33 8.00 57.29
> 129.90 68.20 5.24 100.00
> sdb1 11763.33 0.00 1908.00 1.00 109365.33 8.00 57.29
> 129.90 68.20 5.24 100.00
> sdc 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
> 0.00 0.00 0.00
>
> i have bolded the avrg wait,srvc time and % utilization and i think
> given the above stats the i/o is bad on this configuration. the wait
> times are very high and there is a problem with the way physical
> configuration of the disks is done and how the datafiles are laid out.
>
> please advise me on this and explain whether my observation is
> alright. and also tell me what could be a reason for this to happen.
>
> thanks
> sai
>
> ----------------------------------------------------------------
> Please see the official ORACLE-L FAQ: http://www.orafaq.com
> ----------------------------------------------------------------
> To unsubscribe send email to: oracle-l-request@xxxxxxxxxxxxx
> put 'unsubscribe' in the subject line.
> --
> Archives are at //www.freelists.org/archives/oracle-l/
> FAQ is at //www.freelists.org/help/fom-serve/cache/1.html
> -----------------------------------------------------------------
----------------------------------------------------------------
Please see the official ORACLE-L FAQ: http://www.orafaq.com
----------------------------------------------------------------
To unsubscribe send email to: oracle-l-request@xxxxxxxxxxxxx
put 'unsubscribe' in the subject line.
--
Archives are at //www.freelists.org/archives/oracle-l/
FAQ is at //www.freelists.org/help/fom-serve/cache/1.html
-----------------------------------------------------------------
Other related posts:
