ONTAP Hardware

FAS 3270 HA pair Best Practices

JOHNYTHENMARRY
3,881 Views

  Hi,

I am going to implement FAS 3270 HA pair with four disk shelves and with two types of disks (48 x 600 SAS, 36 x 2 TB SATA) and my plan is to do the configuration based on the below table.

Host Name Aggregate Name RAID     Disk Type     No Data Disk        No of Parity Disk       No of Spare Disk

Filer-A         Aggr0         RAID-DP        SAS              20                        2                                 2

Filer-A         Aggr1         RAID-DP        SATA            14                        2                                2

The same configuration will be applied for the HA Pair also.My design was based on considering the below points.

  1. Maximum RAID Group size is for RAID 6 is (RAID-DP): FC – 28 (26 data disks plus 2 parity disks)
  2. Maximum RAID Group size is for RAID 4 is (RAID-4): SATA – 20 (18 data disks plus 2 parity disks
  3. Maximum Number of LUNs 4096
  4. Maximum Number  of FlexVol® Volumes Up to 500 per controller
  5. Maximum Volume/Aggregate Size 70TB 
  6.     

If anyone is having better idea please advice.

Thanks

Johny

5 REPLIES 5

michaelll
3,881 Views

I would distribute the disks differently, I would put all of the SAS on one controller and the SATA on the 2nd controller, this will allow all of the disks to contribute to the IO needs of the aggregate, in your configuration you would have 2 aggregates for each disk type using only half of the available IO.

JOHNYTHENMARRY
3,881 Views

Hi Michaell ,

Based on your recommendation if i dedicate one controllerfor SAS and another for SATA what will happen if one controller fails? Whether it can take over to the other controller or the IOs will be terminated?

Johny

michaelll
3,881 Views

Hi Johny, you will not lose any I/O during a failover, Netapp has very clever HA logic.  When both controllers are running it is essentially like having 2 systems, each with their own memory, drives, networking, etc.  When an event that triggers a failover to occur, the remaining controller has the ability to takeover the drives and logical structures (aggregates, volumes, etc) and it will them mimic the failed controller.  Normally this transfer happens with very few lost pings, and well within any I/O timeouts of a guest OS.  When running in failover mode the failed controllers IP's, name, and data area all served through the single controller, so there is a possibility for degraded performance IF you happen to be really pushing the system, but hopefully your reseller has sized your unit properly and this is not an issue.  Always test HA before putting the system into production, and have qualified engineer verify the shelves are cabled correctly for mulitpath HA.

JOHNYTHENMARRY
3,881 Views

Hi Michaell

Thanks for you inputs

Johny

AGUMADAVALLI
3,881 Views

Hi Johny John:

this also depends on the protocols you are going to use along with aggregate layout. As said by somebody, have the sata on one controller and sas on other cluster pair.

thank you,

AK G

Public