[CTDB] some basic info how clustering grew up historically in the open source world
obnox at samba.org
Wed Dec 31 10:43:50 MST 2014
On 2014-12-30 at 17:13 +0100, Günter Kukkukk wrote:
> the former thread about "Setting up OCFS2 and VMs" IMO got
> too much focused on OCFS2 - so i would like to start fresh here
> with some more global view in mind.
Compared to setting up the cluster file system,
setting up CTDB on top is (or should be) a piece
of cake. All the complicated stuff (like quorum,
fencing, etc) is delegated to the cluster/filesystem.
> For all who want to read about the historically backgrounds of
> clustering in the open source world, i think this is a good start:
> It does *not* mention CTDB at all,
Well, CTDB's focus is not so much HA but scale-out.
HA is rather a side effect, but if you don't want
all-active clustering of SMB, you can do without ctdb.
Also, CTDB is not a general purpose piece of clustering
infrastructure, but specifically designed to cluster
SMB as provided by Samba. The fact that CTDB also comes
with small cluster-management system stems from the fact
that at the time when CTDB was created (around 2007), there
was no pacemaker around or anything else to manage all-active
services in a cluster. But nowadays CTDB can in principle
also be run in a database-only mode without doing IP failover
and management of samba and friends, and have itself managed
> but (probably) we need to add some
> additional pieces to get CTDB running on a clustered file system...
> (on different distros)
What do you mean here? CTDB has been deployed successfully
on several clustered file systems since its beginnings in 2007.
Examples are GPFS, GlusterFS, GFS2, OCFS2, and (I think) CephFS.
Again, if the File System is set up properly, then setting up
CTDB top is really really easy. At least it was last time I tried...
Cheers - Michael
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 198 bytes
Desc: not available
More information about the samba-technical