[Linux-cluster] gfs2 v. zfs?

[Gordan Bobic]

Jankowski, Chris wrote:
[...]
> It would require development of:
> - GFS2 specific dump(8)
> - GFS2 specific freeze and thaw commands
> - CLVM wide snapshots
> - more efficient DLM
> 
> It certainly is possible to do. Digital/Compaq/HP TruCluster Cluster File System (CFS) built on top of AdvFS had all of these features and much, much more by circa year 2000.

Indeed, I, too, have been noticing that 10-15 years ago we had all sorts 
of technologies that have all but disappeared and haven't been replaced 
with anything nearly as advanced. Personally, I blame the rise of the 
fenomenon of people calling themselves "programmers" because they can 
write "Hello world!" in HTML. But I digress, and it's somewhat off-topic.

On topic - for consistent "snapshots" for backup purposes there is a 
cobbled together solution that can be used for this sort of thing, if 
you have the option of using DRBD. If you have a DRBD setup you can 
unmount the file system on a slave node and drop it out of replication 
completely (make it standalone). Mount the GFS file system with 
ro,lock=nolock, and back it up with whatever your prefered backup 
solution is. Then bounce DRBD completely on the slave to make it catch 
up with the master.

Gordan