Re: Two node storage failover with lvm and ISCSI
Denny Schierz put forth on 2/21/2011 4:20 AM:
> hi Stan,
>
> Am Sonntag, den 20.02.2011, 20:13 -0600 schrieb Stan Hoeppner:
>> It's not clear to me at this point if you need real time
>> file/filesystem sharing or simply manual fail over from a dead host to
>> a backup server.
>
> than it's my fault :-)
>
> I want failover (the second in your words). If node 1 fails, node 2 take
> over the IP and fire up the ISCSI target daemon. There is no cluster
> filesystems involved. There is no need for them.
> Problem is: we have have bunch of disks connected to _both_ hosts. So we
> need raid, if disks fails in the storage, on the top LVM for manage the
> space and export them via ISCSI.
>
> I hope, is more clear :-)
Actually, no, it's not really clear, but muddier. :(
Now you need to describe your storage back end in detail. You first
mention a single iSCSI target LUN that each host needs access to. Then
you mention you don't have RAID but that you need RAID.
In the typical iSCSI storage infrastructure, one has an iSCSI SAN RAID
controller chassis with many disks. These disks are RAIDed by the
controller which also allows the OP to slice up the space of the RAID
into multiple logical drives each exported as a LUN. This is often
called "thin provisioning". Each LUN is then assigned to a client host
(or more than one in your case). The RAID function is an integral part
of the SAN controller.
I'm guessing your setup is different than this or you wouldn't be askig
about RAID. Could you please describe your storage back end?
--
Stan
Reply to: