Please enable javascript, or click here to visit my ecommerce web site powered by Shopify.

Community Forum > Mounting gluster shares

I'm having problems mounting a gluster share from my QuantaStor system. Mounting it works fine on the QuantaStor host, but on another host (Centos), it fails with

Mount failed. Please check the log file for more details.

When I look in /var/log/glusterfs/mnt-gfs1.log, I see this error:

[2014-04-01 22:58:24.350251] E [glusterfsd-mgmt.c:1470:mgmt_getspec_cbk] 0-mgmt: failed to fetch volume file (key:/gfs1)

I researched the error "failed to fetch volume file" and found a number of suggestions, none of which seemed to help. Several people who reported this problem (not with a QuantaStor server) said that their problem was solved by making the client and server versions of Gluster the same. On my QuantaStor, "glusterfs --version" says it's 3.2.5, while on my client it says it's 3.4.0, so it's plausible that this is my problem. I'm unable to find a 3.2.5 client installer for Centos, so I'm wondering if I can upgrade the version of Gluster on the QuantaStor server, and if so, how? Or does someone know another solution to this problem?

April 1, 2014 | Registered CommenterTim Morgan

Hi Tim,
We've got v3.10 coming out this Friday which has GlusterFS 3.4.2 included with it. That should take care of the problem. There were some changes between Gluster v3.2 to Gluster v3.4 to where the meta-data is placed so it's a good idea to upgrade first before doing the deployment. More detail on the changes for v3.10 are available here.
Best,
-Steve

April 2, 2014 | Registered CommenterSteve

Thanks---that may be our solution.

I noticed in the changelog "adds ZFS dataset creation for gluster bricks". What does that mean to me as a user/administrator? In particular, I'm interested in whether there will be an ability to take snapshots of Gluster shares for backup purposes---will we be able to do that?

April 2, 2014 | Registered CommenterTim Morgan

Hi Tim,
Yes, that exactly right. The way were were allocating bricks before was as simple directories which is OK but it doesn't let us leverage some features like snapshots. In ZFS you can allocate a filesystem aka dataset which appears as just a directory but it's logically a separate entity so you can do some neat things with this when we map those to bricks. That includes things like customizing the sync & compression policy, and the ability to do snapshots and such at the brick level. In the v3.10 release we don't have Gluster snapshot support yet but we've paved the way for that so now we can introduce snapshots in an a future release without having to move data.
-Steve

April 2, 2014 | Registered CommenterSteve