switches required?

[dave_richards]

Hi all,

I am playing with a new FAS3220, dual controllers, running OnTap v8.2.  I am quite familiar with OnTap v7 with older hardware, but this is the first time I have played with cluster mode.

I have 2 linux machines (Redhat) with 2 FC HBA's each, directly cabled to the 3220 controllers.  One machine is connected to controller A port 0c, controller B port 0d.  The other machine is controller A port 0d, controller B port 0c.

I have setup LUN's and presented them with initiator groups etc, but the linux box can't see the LUN's.  Below is the output from controller A port 0c, as an example:

                  Node: mira3220-01

               Adapter: 0c

           Description: Fibre Channel Target Adapter 0c (QLogic 2432 (2462), rev. 2, 4G)

     Physical Protocol: fibre-channel

         Maximum Speed: 4

                Status: online

             Substatus: ADAPTER UP

     Host Port Address: ef

     Firmware Revision: 5.8.0

Data Link Rate (Gbit): 4

    Fabric Established: false

Connection Established: ptp

             Mediatype: ptp

      Configured Speed: auto

      Configured State: up

           Switch Port: :

As you can see, 'fabric established' is false and switch port is empty.  Is direct cabling supported or not?

Also, the OnCommand system describes the network interfaces for the vserver as disabled under 'Operational Status', and 'Administrative Status' as enabled.

The other potential problem I could foresee is that the HBA's in the Linux machines aren't compatible with clustered mode, for some reason.  Is this a thing?

Thanks in advance.

[dave_richards]

Linux reports the HBA's as:

qla2xxx 0000:42:00.0:

QLogic Fibre Channel HBA Driver: 8.03.07.09.05.08-k

  QLogic QLE2560 - PCI-Express Single Channel 8Gb Fibre Channel HBA

  ISP2532: PCIe (5.0Gb/s x4) @ 0000:42:00.0 hdma+, host#=8, fw=5.06.03 (90d5)

[GRAEMEOGDEN · Accepted solution]

Hi Dave,

Direct attached FC configurations are not supported in Clustered OnTap. You can direct-attach iSCSI if you have a single node, but a HA pair requires switches. Check the SAN config Guide: https://library.netapp.com/ecm/ecm_download_file/ECMP1196793

For supported HBAs and drivers, the Interoperability matrix is the place to look: http://support.netapp.com/matrix

Cheers

G

[dave_richards]

Added a switch into the environment, and all things are working well again.  Thanks Graeme.

[dave_richards]

It seems switches are not mandatory with a pair of controllers:

     https://library.netapp.com/ecmdocs/ECMP1367947/html/GUID-CE8ADA56-E05B-4398-A7E2-05C7BD816467.html

[joha]

Correct, two-node switchless clusters are quite common. This was made available from Clustered Data ONTAP 8.2

[GRAEMEOGDEN]

Correct me if I'm wrong, but that is for the cluster interconnects, for SAN / data connections switches are still required.