I am trying to make a FAS2552 HA pair to work with a Cisco C240 server by using direct attach connection, I have tried FCoE and iSCSI but cannot get it to work, I have read from Netapp documentation that FCoE is not supported in direct attach but iSCSI should be support, anyone have experience on iSCSI direct attach of HA pair before?
my name is Chriz Ott, I'm working with NetApp as a Principal Architect.
Thank you very much for bringing this topic up, as I have seen this very often.
Unfortunately we do not support SAN configurations (FCP, iSCSI or FCoE) directly attached to FAS storage systems, as we are providing high availablity by using ALUA and MPIO. From the Scalable SAN FAQ:
"Is a direct-attach configuration possible? Answer: No. The failover capability of clustered Data ONTAP requires an active path to a node’s HA partner. Furthermore, in the case of Fibre Channel configurations, direct connections do not support the N_Port ID virtualization (NPIV) facility."
In order to help you setup your environemnt I can recommend i.e. the iSCSI for Windows Express guide:
Unfortunately we do not support SAN configurations (FCP, iSCSI or FCoE) directly attached to FAS storage systems, as we are providing high availablity by using ALUA and MPIO.
This does not explain why direct host connection with iSCSI is impossible. Host having 2 ethernet adapters, each to different controller, and using MPIO over these two adapters. What does not work in this configuration?
FCoE is not possible because it requires management instance which is normally implemented by switches and FC may have problems with NPIV as you mentioned. But I miss the reason why iSCSI should not work.
So you mean iSCSI with HA is not possible, so a vol move between the node will cause the connection to break? Is this only apply to cluster mode? If I use 7 mode is it possible to use iSCSI without problem?
No, in 7-Mode it is near to impossible to get right. The problem is, 7-Mode iSCSI does not use ALUA so it relies on simple IP failover. If you have direct controller connection and controller goes down, IP is moved to partner, but host is still connected to the same controller that is not available anymore. So it cannot access target IP.
But C-Mode is using ALUA where LIFs themselves do not migrate. So host has two (or more) independent paths to two different controllers. I honestly fail to understand how switch connection differs from direct connection here.
I would be interested to know the technical reasons why direct-attach iSCSI in an HA configuration is not supported, even though there are indications that it will/should work.
If I have the following config:
e0c ---> Host_P1
e0d ---> Host_P2
e0c ---> Host_P3
e0d ---> Host_P4
In the above configuration, the host MPIO should see  paths to a LUN.
In the event of a controller failure, where the SVM would only have LIFs on the surviving controller, the host MPIO would detect the failure of  paths, but have  remaining paths (since the NetApp iSCSI LIF's can only be configured as "Failover Policy: disabled", they can't migrate from the failing controller to the surviving controller).
Previous posts indicate that "it will work, but is not a supported config"... so, I am interested in the technical issues that led NetApp to not support the configuration, even though it will/should work.