[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [Linux-cluster] Replication for iSCSI target?

Customer’s requirements drive the solution, there are many ways to achieve HA storage.  I don’t know your exact setup and resources but I reckon iSCSI will be just an unnecessary overhead layer if you are using GlusterFS. You can use DRBD with a clustered file system (e.g. gfs, ocfs, ..) or with  xfs/ext3/4 + NFS, you may consider using G/NBD as well. Again, the solution is driven by the customer’s  requirements and the available resources.  GlusterFS is an awesome solution, provides HA & performance but it has some limited capabilities too, like it doesn’t support POSIX ACL. Lastly my 2cents for a clustered file system, it sounds like a sexy solution but if it’s deployed in a small scale (e.g. < 5 nodes) it might easily become a nightmare. For small scales HA storage without losing stability I’d recommend DRBD + XFS/EXT3/4 + NFS.



From: linux-cluster-bounces redhat com [mailto:linux-cluster-bounces redhat com] On Behalf Of chenzp
Sent: Sunday, 3 April 2011 2:43 PM
To: linux clustering
Subject: Re: [Linux-cluster] Replication for iSCSI target?


use drbd!


2011/4/2 Michael McGlothlin <michaelm plumbersstock com>

I'm experimenting with setting up an iSCSI target that has data
replication between three nodes. Right now I'm trying Glusterfs which
seems workable but I'm not sure how it'll handle it if more than one
node is trying to access the same target device (1TB sparse file) at
the same time. Has anyone set something like this up before and can
give me some hints? I was looking at GFS before but it appeared to not
do replication? DRBD seemed like a possibility but having more than
two nodes sounded as if it might be an issue.

Each server has dual quad-core Xeon processors, 64GB RAM, 8 2TB
drives, and 10Gb Ethernet so I hope hardware won't be a limitation.
We've constantly had trouble with every iSCSI, SAN, NAS, etc we've
tried so I want to make something that is completely void of any
single point of failure.

Michael McGlothlin

Linux-cluster mailing list
Linux-cluster redhat com


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]