Unix and Linux → Gluster FS

Hello,

I am looking for a way to add redundancy, HA and flexibility to expand and reduce volumes. Although, i have an EVA4400 that is pretty redundant, managing volumes for VMs is kinda of a pain.

I am looking for something like Microsoft DFS for Linux. I came across Gluste FS which seems to be pretty good, in fact they were bought by Red Hat.

I would like to be able to have my linux servers to store data in the same fashion as in MS DFS. Where you access the share like "\\DOMAIN.LAN\FOLDER" so far the documentation i've read on Gluster all points to "\\SERVER\FOLDER" which kinda of contradicts itself...

Anyone has any suggestions..?

Thanks

We evaluated Gluster a few years ago but went with hadoop, so my memory is 'fuzzy'.

But, as I understand it, Gluster uses fuse as a mount point into the cloud to give posix file operations. You're directly accessing the storage nodes, fuse is doing that via the mount.

So basically, there is no single/generic mount point that clients can access, correct?

It would be nice to have something like MS DFS, you can access shares via a single domain name and the DFS takes care of the rest.

Thanks for the info.

Each client has it's own virtual mount via the local fuse install (again this is all two year old recollection).

You really don't want that mount to be centralized else every request has to go through that hardware resulting in a bottle neck. You want each client pulling blocks from the cloud automously so the load is spread across the servers providing the blocks.

Technically you could make a centralized fuse mount and then put an another networked mount pointing to it, but you lose performance, and likely security by doing so.

Hi there,

You mount GlusterFS much like you would an NFS or CIFS share. Once you mount it, you treat it as you would any other directory.

With GlusterFS, the shares can be distributed, replicated, distributed + replicated, geo-replicated (asynchronous)

If you're accessing the shares via Windows, I would actually recommend you use the object storage piece, bypassing CIFS and NFS altogether. The object storage (a la Amazon S3 or Swift) is a feature of the 3.3 beta, but there's a backport available for 3.2.x. From the Windows client side, there are tools you can use to easily access the share via object storage. Cyberduck is one such tool - »cyberduck.ch/ - but there are plenty of others: »www.dragondisk.com/ , »www.cloudberrylab.com/ , etc.

For details, feel free to ask on gluster-users@gluster.org , »community.gluster.org/ or #gluster on irc.freenode.net

-JMW
GlusterFS community guy