Re: questions about weight

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi Ravi,

On Sun, 2010-11-07 at 23:52 -0800, Ravi Pinjala wrote:
> First off: what does weight actually mean in a CRUSH map? I had
> assumed it was there to balance capacity, and should be set to
> something proportional to the size of the disk. However, the page on
> the wiki about osd_auto_weight seems to indicate that it's related to
> disk performance, not size. Is it one of these, or something
> completely different?

It is the size of the disk, for example:

osd1, size 120GB, weight: 1.200
osd2, size 560GB, weight: 5.600
osd3, size 340GB, weight: 3.400

Now, you should be careful, since osd2 and osd3 will have to do much
more IOps, so make sure their hardware is up to the job.

> 
> Also: when defining several levels of buckets, how does the weight on
> the higher levels of buckets affect the overall distribution? I've got
> the beginning of a cluster running now, with one host with a 640GB
> drive and a 1TB drive, and another host with a 640GB drive. I've been
> playing with the weights to try and get the amount of data on each
> disk proportional to its size, to maximize the space I have available.
> I have my CRUSH map set up like this right now:
> 
> host0 {
>     item device 0 weight 0.640
>     item device 1 weight 1.000
> }
> 
> host1 {
>     item device 2 weight 0.640
> }
> 
> root {
>     item host0 weight 1.640
>     item host1 weight 0.640
> }
> 
> Will this do what I'm expecting, and maximize the amount of data I can
> store before any of the disks fill up?

You don't have to calculate a hosts weight in the root, this will be
done automatically by summing the device's weight.

But, it looks fine what you've done, that should work.

Wido

> 
> Thanks,
> --Ravi
> --
> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html


[Index of Archives]     [CEPH Users]     [Ceph Large]     [Information on CEPH]     [Linux BTRFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]
  Powered by Linux