[Linux-cluster] Re: Storage Problems, need some advice

[Kevin Anderson]

We use iSCSI in our development labs for GFS clustering.  Basically, GFS
requires that the underlying storage be concurrently accessible from all
nodes in the cluster.  This rules out a shared scsi bus between nodes,
but fibre channel, iSCSI and gndb all provide the ability to access the
underlying storage concurrently.   From the basic descriptions of the
DS300, I don't see any reason it wouldn't work very well with GFS.  It
provides either fibre channel or iSCSI support.

Kevin

On Tue, 2006-11-07 at 18:33 -0500, David Brieck Jr. wrote:
> On 11/7/06, Riaan van Niekerk <riaan at obsidian.co.za> wrote:
> > > After doing much more research I came across the DS300 by IBM. It uses
> > > SCSI drives, is fully redundant, does iSCSI and doesn't cost an arm
> > > and a leg (just an arm). My question is, their site says linux
> > > clustering isn't supported, but does it have to be? Doesn't iSCSI let
> > > you do the same thing GNBD does?
> > >
> >
> > hi David
> >
> > do you have a link to the page with that statement? "linux clustering"
> > is somewhat of an ambiguous term. Within the context of Red Hat software
> > (excluding Linux Virtual Server and high-performance computing
> > clusters), it can mean either:
> >
> > a) Cluster Suite (without GFS) - only one node in a cluster accesses the
> > storage at a time. if you fail/switch over, one node unmounts an FS,
> > another one mounts it.
> >
> > b) GFS (which implies/includes Cluster Suite) - multiple nodes accessing
> >    the same LUN with a (G)FS on top of it
> >
> > I am not familiar with entry-level iSCSI initiators. I always thought
> > iSCSI is logically like fibre, e.g. multiple hosts in the same raidgroup
> > can concurrently access the same LUN/FS. Perhaps these entry-level iSCSI
> > arrays are more like regular SCSI meaning that they do not support
> > multiple initiators accessing the same LUN behind a target (storage
> > processor).
> >
> > I had a look at EMC cert matrix for the AX100/150 series arrays
> > http://www.emc.com/interoperability/matrices/AX_Series_SupportMatrix.pdf
> > thes entry-level EMC iSCSI arrays also only supports non-clustered Linux.
> >
> > iSCSI will allow you to "do the same thing" as GNDB:
> > GNDB client and server are replaced iSCSI initiator (Linux host) and
> > target (dedicated hardware, e.g. EMC array, or software target - not yet
> > considered production-ready nor included with RHEL). However, if the
> > hardware has an explicit exclusion of Linux clustering, you are stuck,
> > not being able to have two nodes aaccess the storage at the same time..
> >
> > HTH
> > Riaan
> >
> > > Also, I talked to someone on their chat who said I could use any U320
> > > drive with it, basically I could reuse the drives I already have and
> > > just not use my old enclosure. Does that sound right? Any reason I
> > > couldn't do that other than loosing all my data?
> > >
> > > Anyone using a DS300? Seems like with 15k drives it would be pretty darn
> > > fast.
> > >
> > > --
> > > Linux-cluster mailing list
> > > Linux-cluster at redhat.com
> > > https://www.redhat.com/mailman/listinfo/linux-cluster
> >
> >
> > --
> > Linux-cluster mailing list
> > Linux-cluster at redhat.com
> > https://www.redhat.com/mailman/listinfo/linux-cluster
> >
> >
> >
> 
> Here the PDF I was looking at:
> http://www-03.ibm.com/servers/storage/disk/ds/pdf/ds300400_interop.pdf
> 
> It says it supports Microsoft Clusters so I don't know why it wouldn't
> support linux. I also found somewhere else that said more specifically
> that it doesn't support RedHat Cluster Manager, but as far as I'm
> concerned it has nothing to do with that unless you're talking about
> fencing, in which case it would work for me since I'm using power
> fencing.
> 
> Anyone from RedHat mind chiming in? It's not on the hardware
> compatibility list, but there are many things that work that aren't on
> that list.
> 
> --
> Linux-cluster mailing list
> Linux-cluster at redhat.com
> https://www.redhat.com/mailman/listinfo/linux-cluster
> 
> 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/linux-cluster/attachments/20061107/90c33b60/attachment.htm>