2013-04-10 09:26 PM
Got some Windows 2008 R2 hosts using Microsoft iSCSI initiator to connect to NetApp storage. Storage controller has a single-mode ifgrp with two VLANs for the host to connect to. The host has two NICs, one on each VLAN. The two host NICs and the two controller interfaces are balanced across two switches (no ISL). Because only one of the ifgrp interfaces is active, there is effectively one active path from the host to the iSCSI target, via one of the two VLANs over one of the two switches. See attached diagram.
This morning the ifgrp active NIC changed, possibly due to a network or hardware issue. Since around this time hosts have been having trouble connecting to their LUNs. Microsoft iSCSI initiator appears to be OK, it can see the target IQN and storage devices, but DIsk Manager and Device Manager are not happy.
Would you expect the act of switching active interfaces in the ifgrp would have any impact on the iSCSI host? Effectively it would cause a path failover for the host LUN, from VLAN "100" on NIC "a" via switch "a" to VLAN "101" on NIC "b" via switch "b" - is this expected to be disruptive?
2013-04-10 10:35 PM
Changing of active interface in single mode ifgrp should have no impact on host and should not cause path switch from the host point of view. Host continues to see the same target IP over the same local interface.
Your configuration is invalid. Switches to which two interfaces of ifgrp are connected must have level 2 network connectivity for all VLANs on this ifgrp. I’m surprised you were able to bring ifgrp up at all; Data ONTAP should verify connectivity between ifgrp members.
2013-04-10 10:59 PM
Switches to which two interfaces of ifgrp are connected must have level 2 network connectivity for all VLANs on this ifgrp
Yes I noticed that from the Network Management Guide, this is now in the process of being rectified.
In the current configuration the target IP does change, would this normally have an impact on the host?
2013-04-10 11:28 PM
In our case... single ifgrp members are all in the same vlan. So our target IP doesn't change in case of failover.
It's best to test your "single ifgrp" with the "ifgrp favor" cmd.