#Do I Need Intercluster LIFs on FAS System for ONTAP S3 with FabricPool setup

1 messages · Page 1 of 1 (latest)

agile nymph
#

Hello All,
I am in the process of configuring ONTAP S3 on a FAS system and plan to use it for FabricPool with an AFF system. I am a bit uncertain about the network configuration requirements.

Do I need to create intercluster LIFs on the FAS system where ONTAP S3 is configured, or can I use ONTAP S3 data LIFs to connect to the AFF system for the FabricPool setup?

Any insights or recommendations on the correct approach and best practices for this configuration would be greatly appreciated.

Thank you!
Vast

wintry sky
#

These are two clusters right?

#

On the source cluster (performance tier) you need 1x intercluster LIF per node.
On the destination cluster (capacity tier) you need 1x S3-data LIF inside your S3-SVM.

#

Peering is not needed, you don't need intercluster LIFs on your destination

#

These LIFs will be used for the FabricPool traffic so best put them in a separate VLAN if possible.

agile nymph
#

Yes, these are two separate clusters: one AFF (performance tier) and one FAS (capacity tier).

Should the VLAN be configured as Layer 2 on both clusters, or is it necessary for them to be Layer 3?

pulsar flume
#

Yeah, you need S3 data LIFs (create a matching service-policy with data-s3-server first) on the destination SVM, and you need intercluster LIFs on the AFF that is tiering the data out
Layer 2 or Layer 3 (routed or not) does not really matter, unless your router is slow 😉

mystic raven
#

Honestly you might be better off making one cluster (if you can get the switches) then you don’t need to worry about certificates and intercluster lifs

#

Lots of customers do this.

#

I think it’s even one of the specific examples in the docs when configuring

pulsar flume
#

Most people I talked to discourage intra-cluster FabricPool though, for the simple reason that the whole S3 traffic runs over the cluster LIFs which can impact the cluster backend performance and cause all kinds of trouble

wintry sky
#

With 100Gb Cluster network that shouldnt be much of an issue.

pulsar flume
#

with 40G it apparently still was a problem, at least according to some engineers at NetApp 🤷‍♂️

pulsar flume
#

also you would need Cluster switches, which make the total setup a bit more expensive 😉

mystic raven
#

If you are using 40/100g switches with newer platforms then ONTAP will utilize RoCE on the switches for improved performance

#

Of course the rcf/code needs to be up to date

#

Great benefit from the 1.0 versions to the 1.8