Home / server / Questions / 138380
In Process
w00t
Asked: 2010-05-05 05:47:13 +0800 CST

Solaris to Linux conversion: Use VxFS or GFS?

  • 772

We're a Solaris shop looking at RedHat Enterprise Linux and one of the things we're wondering is if we should keep Veritas Volume Manager + FileSystem or go with LVM+ext3 or RedHat's preferred cluster filesystem solution, GFS.

One of the things we like about Veritas is that it can use Veritas Volume Replicator to have a remote copy of important filesystems. This functionality seems to be missing from RedHat, DRBD doesn't seem to be packaged in RHEL...

So my questions are:

  1. Does anybody use VxFS/VxVM/VVR on Linux? Thoughts, experiences? Comparison with LVM+ext3?
  2. Anybody using GFS? Thoughts, experiences?
  3. Do you do remote replication for disaster recovery, and if so, how? Is there a standard RedHat way?
redhat replication veritas rhel5 gfs

5 Answers

  1. My answers:

    1. We have used VxFS and VxVM (but not VVR) on Linux. They work just as expected and exactly the same as under Solaris. No big surprises there. As for comparison with LVM/ext3, that's all personal preference. Both combinations are solid.
    2. Not using GFS currently, however, it is more in line with Veritas Cluster File System where you would have two servers with concurrent read/write access to a shared file system. I wouldn't see its being used as a DR/replication product.
    3. Most of our replication is handled on the back-end storage arrays, so I've no answer for you there.
    • 2

  2. Suse supports DRBD. I've used DRBD (over debian) for remote replication with no mayor issues. You have to tune the protocol according to your bandwidth.

    • 2

  3. If you're going for RHEL, why not contact Red Hat with these questions?

    • 1

  4. Personally, I've not used or seen VxFS under Linux. Under Solaris, it seems to be pretty solid, though.

    For clustering, I would look at some of the suggestions mentioned in answer to a question I asked: GlusterFS, Lustre, GFS, Hadoop, GPFS, etc.

    Also note: with the release of Red Hat Enterprise Linux 6, standard support for ext4 and XFS will be included.

    • 1

  5. We've seen some nasty problems around locking with GFS2 (eventually requiring a re-boot of all nodes in the cluster), and, to a lesser extent, GFS when used across multiple nodes (we have open tickets with Redhat).

    GPFS from IBM has been fine, as has CFS from Veritas, but unfortunately both these options cost money :-)

    Having said that, we do seem to have an unusual application which uses locking intensively. I imagine the vast majority of systems are not seeing the issues we are. So, as always, 'suck it and see' :-)

    • 1