[Lustre-discuss] testing block device performance while lustre is running

Brock Palen brockp at umich.edu
Tue Dec 14 12:40:16 PST 2010


On Dec 14, 2010, at 3:37 PM, Joe Landman wrote:

> On 12/14/2010 03:33 PM, Brock Palen wrote:
>> Is there a safe way to test block device read performance while
>> lustre is using the device?  Example a type of read only access to
>> the device to check that it is performing correctly?
> 
> Brock
> 
>   We allocate a small aligned partition before the Lustre OST partition 
> for read and write testing.  If you don't have this, you might be able 
> to do read testing.  The latter you would need to do with dd on the raw 
> device
> 
> 	dd if=/dev/sdXP of=/dev/null bs=128k count=Z

This is what I thought I could try, we have just seen a degration in performance on our OST's and trying to track down why. 
Thanks

> 
> where X is the drive, P is the partition number (if any), and Z is the 
> number of 128k blocks you want to read.  Don't mix up if and of 
> arguments (e.g. this can be a dangerous command with a minor typo)
> 
>> 
>> Or is this a bad idea?
> 
>   It will impact OST performance if you wind up increasing resource 
> contention.

Figured
Thanks

> 
>   Regards,
> 
> Joe
> 
>> 
>> Brock Palen www.umich.edu/~brockp Center for Advanced Computing
>> brockp at umich.edu (734)936-1985
>> 
>> 
>> 
>> _______________________________________________ Lustre-discuss
>> mailing list Lustre-discuss at lists.lustre.org
>> http://lists.lustre.org/mailman/listinfo/lustre-discuss
> 
> 
> -- 
> Joseph Landman, Ph.D
> Founder and CEO
> Scalable Informatics Inc.
> email: landman at scalableinformatics.com
> web  : http://scalableinformatics.com
>        http://scalableinformatics.com/sicluster
> phone: +1 734 786 8423 x121
> fax  : +1 866 888 3112
> cell : +1 734 612 4615
> _______________________________________________
> Lustre-discuss mailing list
> Lustre-discuss at lists.lustre.org
> http://lists.lustre.org/mailman/listinfo/lustre-discuss
> 
> 




More information about the lustre-discuss mailing list