Gluster users - Apr 2017 - Is there any limit on the number of volumes

Hi,

We are planning to use GlusterFS for an application that can potentially create
about 100,000 volumes (each of quite small capacity).
At any time, only about a 500 of them could be mounted across multiple compute
nodes.

Is there any limit on the number of volumes a GlusterFS cluster can have?

I would also like to know:
- if there is any performance impact of having a large number of volumes
- do we need to provision any resources differently for this

Thanks & Regards,
Tanmay.


-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.gluster.org/pipermail/gluster-users/attachments/20170404/0aeed4b5/attachment.html>

Few questions inline:

On Tue, Apr 4, 2017 at 2:36 PM, Tanmay Mohapatra <tanmaykm at hotmail.com>
wrote:
> Hi,
>
> We are planning to use GlusterFS for an application that can potentially
> create about 100,000 volumes (each of quite small capacity).
> At any time, only about a 500 of them could be mounted across multiple
> compute nodes.
>
More than the mounted volumes (or number of mount points), we do check our
resources in terms of brick processes. So, a better question to answer
would be how many volumes will be in started state at any point of time.

Again, what is small capacity volume for you? Gluster as of now treats
'brick' as a resource (ie, hostname and an export directory) along with
a
'volume' which contains one or more 'bricks', as another
resource.  So
number of bricks per volume also matter for defining the max number of
volumes.


>
> Is there any limit on the number of volumes a GlusterFS cluster can have?
>
> Officially we haven't declared a limit, available memory and other disk
resources would matter to define these limits. We have tested with 1000s of
volumes, but 100k volumes is not tested till now IMO.

> I would also like to know:
> - if there is any performance impact of having a large number of volumes
>
Yes, gluster CLI commands  *may* become slow. It doesn't matter for
filesystem interface.

> - do we need to provision any resources differently for this
>
>
I don't have a correct answer at the moment.

> Thanks & Regards,
> Tanmay.
>
>
>
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://lists.gluster.org/mailman/listinfo/gluster-users
>


-- 
Amar Tumballi (amarts)
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.gluster.org/pipermail/gluster-users/attachments/20170406/0c656d34/attachment.html>