[Gluster-users] Recommendations for busy static web server replacement

[Liam Slusser]

answers inline

On Wed, Feb 8, 2012 at 1:09 PM, Brian Candler <B.Candler at pobox.com> wrote:
> Interesting - how do you achieve the 'breaking' of the pair? Do you just
> create new distributed volumes containing the same bricks but only from one
> side?

Yes, I create new distributed volumes containing the same bricks but
only from one side.

> I think they may be trying to prevent you doing this in future:
> https://github.com/gluster/glusterfs/commit/cf944c8ad5da87bce15b08d0bbb2ecd62e553d86
> but I'm sure you can get around it with symlinks or something.

Might be the case.  But it works great with Gluster 2.x, I haven't
upgraded to 3.x yet.

>
>> e) I have a very small 2mb cache in our gluster clients.  We have such
>> a large volume/library that getting a cache hit almost never happens
>> so don't waste the memory.
>
> How is that tuned? Is there a mount option for it?

Its specified in the glusterfs.vol client configuration file under
"volume cache".  However I do use a readahead cache.

\> Does that mean: you're exporting the same filesystems as different bricks?
> (Otherwise I can't see how you bind the different ports)

No, exporting the filesystem as the same brick names.  But I run
multiple copies of gluterfsd (the server daemon) with different
configuration files pointing to the same bricks.  Inside that
configuration file I am specifying the port number.

So on the server...

gluster running on port 6996 exports brick1a=/brick1a and brick1b=/brick1b
gluster running on port 6997 exports brick1a=/brick1a and brick1b=/brick1b

I then use different client configuration files to mount the bricks

On the client...

/mnt/gluster0 = /home/gluster/gluster0.conf which mounts server
192.168.0.50:6996 brick1a and brick1b
/mnt/gluster1 = /home/gluster/gluster1.conf which mounts server
192.168.1.50:6997 brick1a and brick1b

> :-)
>
> Regards,
>
> Brian.

liam