strickt redundancy/strickt failover(active/passive) or active/active operation?

[kerim mohammed]

  Hi,

i couldn't find a documentation which states clearly how traffic flows from the mezanina adapter ports to FEX adapters. We have the A side and the B-side. that is for both fabric interconnecto and  FEX adapters. does the B side actively send and recieve traffic(active/active) or is it a redundant setting (replica of side A) waiting idle till something goes wrong with side A (active/passive). if it is active/active, how is loadbalancing algorithem working at server adapter to pick say the B side as opposed to the A-side? please put it clearly. all the discussions i see talk about how traffic is forwarded on the A side and they explain failure scenarios which only makes sense when the B-side is just redundant and standby. please let me know.

Thanks,

Kerim

[abbharga]

Hi Abdul,

Both the fabric (A & B) are active for both eth and storage traffic at the physcial level. The load balancing can be done / decided by the OS e.g. with vmware you can have the vswitch presented with the NICs and then decide at the vswitch level on how to use these NICs.

However the "failover" feature which you can enable on the NICs at the hardware level, while creating them will have one path as active and will failover to the otherside (at the hardware level) in case the other fabric goes down.

Hope this helps!

./Abhinav

[kerim mohammed]

Hi Abhinav,

thanks! this made it a bit clearer.you saied both fabrics (A & B) are active/active physically and not logically.let me see if i get this right. for a mac address or veth port which is currentlyt on fabric A to be seen on fabric B, there has to be failure on fabric A and vice versa.

the other thing i read on a documentation that for every vEth port on FI(fabric Interconnect) A, a replica will be created on FI-B which will be just sitting there waiting for failure on fabric A. is this true? if so is the reverse also true i.e for every active vEth port on FI-B a replica will be created on FI-A waiting for failure on fabric B. what does this say about mac ddress table size about the FI? please let me know.

thanks,

Kerim

[abbharga]

Hi Kerim,

Let me write a few more things here which should give you some more idea:

1) The failover is not enabled by default for all the vnics you create. You need to do it while creating the SP,

2) Once you have the failover enabled - is when an active interface is created on the main fabric and the standby or passive is created on the other FI.

3) However this hardware level failover is not suggested / avaiable depending upon the adapter type or the OS.

Hope this helps!

./Abhinav

[kerim mohammed]

Thanks Abhinav! that cleared it up! i appreciate the help. talk to you soon.

Kerim