Hi. This probably possible to do with GlusterFS, but I'd like to get the experts opinion. I have several servers running, each writing to their own disks. Also, I have another server that needs to access this info. What I'd like to do, is to unify all the servers disks into a single volume, without HA/tolerance (for now). The the server could access and read the information from this volume in a centralized way. So I interested in following: 1) Re-use existing data - without reinstalling the servers. 2) Export all into a single volume. 3) Potentially in future, add fault tolerance by adding more servers/disks. I asked before about 1 and 2 - but would like to be sure that it indeed would work. 3 also would be nice. A simple sample file for such topology won't hurt as well ;). Thanks. -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20090209/f2dca9bf/attachment.html>
> > This probably possible to do with GlusterFS, but I'd like to get the experts > opinion. > > I have several servers running, each writing to their own disks. Also, I > have another server that needs to access this info. > > What I'd like to do, is to unify all the servers disks into a single volume, > without HA/tolerance (for now). The the server could access and read the > information from this volume in a centralized way. > > So I interested in following: > > 1) Re-use existing data - without reinstalling the servers. > 2) Export all into a single volume. > 3) Potentially in future, add fault tolerance by adding more servers/disks. > > I asked before about 1 and 2 - but would like to be sure that it indeed > would work. > 3 also would be nice. > > A simple sample file for such topology won't hurt as well ;).Have you looked at NUFA? The latest NUFA and distribute translators are compatible with each other. You can have the 'writers' in a NUFA cluster (where they write to local nodes, but with a global namespace -- everybodys data accessible). With the same volfile, you can replace nufa with distribute, and have external machines access this volume as well. Thanks, Avati
Hi. Have you looked at NUFA? The latest NUFA and distribute translators> are compatible with each other. You can have the 'writers' in a NUFA > cluster (where they write to local nodes, but with a global namespace > -- everybodys data accessible). With the same volfile, you can replace > nufa with distribute, and have external machines access this volume as > well. >This sounds as what I look for. So, this approach allows me to have the servers each writing locally to itself, and exposing the data into a single volume. And because it's NUFA, I can add later on replication between the servers in P2P fashion? Regards. -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20090214/6ce396fc/attachment.html>