FC and failover?

[blindzero]

Hi everbody,
we are using two Open-E DSS V6 so far with iSCSI replication and failover.
As we are using the infrastructure now for more and more applications we think about moving to FC with the platform, mainly regarding performance / TCP overhead issues.

As I think about the new platform some questions come into my mind:

1. FC replication?
   At the moment the replication between node #1 and #2 is done by a GbE dedicated cross-cable.
   
   • Is it possible to use some FC connection for the replication as well or is that only possible by GbE and I have to use 10GbE or even bonding?
   • Is bonding a solution at all because the replication session is created once and bonding is not giving any performance boost?
2. FC failover
   For my understanding FC is only another way of connecting the server nodes to my storage. At the moment I have two iSCSI interfaces on each storage node, used with MPIO. I would use the same style to have two redundant network paths from the servers to my storage nodes. But how to handle failovers?
   
   • Is there any FC failover configuration on the Open-E? I couldn't find any.
   • Do I have to implement this failover completely on each server FC HBA configuration? That seems to be some kind of ugly to me regarding all failback, STONITH, etc...

Thanks for your input and clarification to that some kind of new technology for me ;-)

Thanks,

Matthias

[gharding]

Hi,

1) replication is only across ethernet, use either a 1Gb network card or a 10GBe card. We are using a 1GB ethernet card and have no problems with replication and we have 24 vitrtual servers running on 4 ESXi hosts connected to open-e.

2) Failover is a manual process, if the Active server fails, you have to change the secondary servers FC volumes to source and added them to the fibre channel Groups, added them in the same order as on the failed server i.e LUN id. On the ESX server you have to rescan the Fibre Channel Cards and you may have to rename/signature the volumes not sure about if the Volumes where on a native windows/linux server.

It would be nice to have
1) Replication via FC
2) Auto Failover

Regards

Gavin