FAS2040 Good Practice Configuration

[dtrujillo63]

Hello All, I've been setting up my new netapp equipment for the last week and have a few questions. The first thing I did was upgrade to ONTAP 8.0.1 and destroy the 32bit root aggregates to make room for 64bit root aggregates. I have also been playing around with the common functions, LUNs, CIFs, Deduplication, Snapshots, etc. So far everything looks good. Before I go live, I have a few pending good practice questions. My production FAS2040 is configured as an active/active system, the first controller will handle nothing but iSCSI traffic (VM infrastructure, Exchange, SQL, ect) and the second controller will handle nothing but CIFS traffic. This is what I've been thinking for port configuration, please let me know if it will work: configure three Ethernet ports on each controller to handle their primary traffic and configure one port to handle partner failover traffic. For example, the iSCSI controller will have three ports on the iSCSI network and one port on the CIFS network. The CIFS controller will have three ports on the CIFS network and one port on the iSCSI network. I do understand that my environment will be degraded should one controller fail, I could live with that for a few hours. The second topic I have questions about is port grouping. Should I bundle the three traffic ports on each controller to a group and create a virtual interface? Would that provide me with multi-pathing and load balancing? The only thing I have configured on the switch side is jumbo frames, should I make any other changes or adjustments? The filer is connected to two cisco 3750s.

Part of this project also includes a second FAS2040 for DR. The one question I have for that system is regarding an active/passive configuration. With this system only having 12 internal drives, I do not need an active/active system. Is there a way for me to configure the system in an active/passive configuration and eliminate the root aggregate from the passive controller? The reason why I'd like to do this is to eliminate losing three drives for the passive root aggregate. Would the passive controller still takeover if it doesn't own any drives nor have a root aggregate/volume?

Thanks!

[roman_verysell]

Unfortunately you can't skip drives from secondary controller, but, if you (really) want, you can skip clustering completely, with switching to single controller configuration (better way is to order single controller system from NetApp, otherwise you will need to pay doudle for sw licenses for each of controllers).

Also you can make an extremely asymmetrical configuration, with a 2 drives min (RAID-4 w/o spares. It's not recommended and "bad practice", but it works with some voodoo magic) on a quasi-passive controller.

> Would the passive controller still takeover if it doesn't own any drives  nor have a root aggregate/volume?

No. If controller is working, then it need to boot from it's own root. No root - no boot |-)

[mbonneau79]

I would probably do 2 + 2 as far as the interfaces on each of the filers, 2 ports for the primary role and 2 ports for the failover role. 2x 1GBe can handle a fair bit of load, and then you'll never second-guess your performance in a failover situation.

As for DR, I would definitely look at a single head (non-clustered) configuration for your DR filer. I run an active/active 3140 in production with 8x 300GB shelves and a single head 3140 with 4x 1TB shelves for DR. In a failover situation, we have no noticable impact, and in a DR situation it's enough to keep us up and running until the primary can be restored.