Redundant iSCSI SAN

I was checking the features list for something like failover but couldn't find something like that. A through the forum on failover gave me this http://forum.open-e.com/showthread.p...light=failover topic...
In here there's an anouncement for july/august 2006 for automatic failover...is it already there and if not when will it be available ?

For my iscsi solution I had two types of setup in my mind, depending on what will be possible, could someone tell me if they're possible :

Setup 1
- Two storage servers with local disks.
- Both nodes export half of the total available lun's to spread the load.
- Both nodes replicate their active lun's to the other node to keep in sync.
- In case of failure the still active node exports all lun's. (possibly automatically)

Setup 2
- One or more FC storage arrays with redundant controller
- Two DDS storage servers as FC > iSCSI gateways
- Both nodes export half of the lun's that reside on the FC arrays, to spread load.
- In case of failure the still active node exports all lun's. (possibly automatically)

In the second case I use Open-E as a storage gateway to virtualize underlying arrays and to provide snapshotting and a decent gui for management.
For both setups automatic failover would be very nice...manual failover in my case means all xen rootfilesystems running from SAN will be dead first and all vm's have to be restarted.

I was also wondering how the manual failover procedure works and if it's possible to script it from the outside ?

Regards,

Werner Reuser
XL-Data Hosting

We did not complete it then and hopefully by same time this year Aug. we will have it. This is a very complicated feature and we will not release until it is sufficiently stable. So currently in your case solution you will need to set up 1 Source server and 1 Destination server for failover procedures, this is done manually (currently) and we do not have a script to run from the outside. So if Source fails then you instruct Destination to be Source and enable the Target (See manual for these steps) - then instruct initiators to use the new Source IP address. If old Source comes back on line set it to either Destination (See manual for deeper explanation) or Set New Source back to Destination and failed Source as Source then reset the Metadata - clearing of metadata is required when you want to start the replication process from the beginning.

I understand it is a difficult feature...but to me it seems as a basic requirement...You can't run primary storage on a SAN that isn't redundant. Reading the features list and the manual I also can't find any failover options for nfs / cifs...those two aren't really hard to implement.
We build NAS solutions for ourselves using DRBD for replicating and Heartbeat + Stonith for the failover...pretty easy setup...

The manual procedure could be an option to get online again quickly...but in the mean time all my iscsi connections get killed and all virtual machines are killed as well...
I guess I'll have to wait for the failover version then...

If possible I would like to (beta) test the (iSCSI) failover version...

We provide the beta versions on our FTP site located in the Engineering phase directory. Usually our resellers test these but you have access to this as well. When testing send all information and results to support@open-e.com. Please keep in mind though we have so many responses that we can get overwhelmed with reports and really don’t want to advertise this to much for obvious reasons. So keep an eye on it!! :)

Hi To-M,

I emailed support for a login / url...

Thnkz !