2014-08-10 11:19 PM
I am very new to netapp , currently in the process of setting up its network connectivity. The setup would be done with multiple VLANS in such a way that , controller-1 would be dedicated for virtualization (using ISCSI/NFS) and controller-2 would be for NFS/CIFS file shares.
I have a quick question regarding connecting a CISCO 3570 X stack to FAS2240 with dual controller. Which is the recommended way ? I have some assumptions but not so sure. Any help would be really appreciated.
1. Out of two interfaces of controller-1 , connect to each switch in the stack (say a stack of CISCO 2 x 3570 X series). Similarly for controller-2.
2. Two interfaces of controller-1 to first switch in the stack and two interfaces of controller-2 to second switch in the stack.
In both cases, should we have to choose a partner interface while configuring VIFs?
Thanks in advance for help.
Solved! SEE THE SOLUTION
2014-08-11 12:00 AM
Each NetApp controller is independent and has own connectivity. So variant 1 will give high available connection for each controller. And yes, to ensure addresses are taken over by partner during controller failure you need to set partner interface.
2014-08-11 01:24 AM
Thanks a lot for your reply. I'll go with 1st variant...
However in variant 2, suppose if the switch-1 which is connected to controller-1 fails, are we lost with controller-1 connectivity completely ? Can partner interface in controller-2 takes over the addresses and route traffic via switch-2 ?
2014-08-11 01:43 AM
Yes, NetApp supports failover on loss of connectivity. Check documentation for “negotiated failover”.
To avoid misunderstanding - it is not failover of interface. It is failover of complete controller - i.e. one controller is shut down and all services are migrated to partner.
2014-08-11 02:18 AM
So to summarize, if I go with variant2, with "negotiated failover" enabled, still we can survive with one switch failure, provided the whole load would be on the second controller!.
In variant 1 since its dual connected on each switch in the stack, we have the controller available without a partner migration being triggered. Which is exactly would be apt for my need as we could dedicate controllers for different purposes. And at last a partner interface config would help in the event of controller failure.
Thank you again.