#Can this configuration be supported?

1 messages · Page 1 of 1 (latest)

dire sonnet
#

We've an existing configuration: 2 x A300 with 2 x 10G on each node and 2 x C60 with 100G on each node for CISCO N9K cluster interconnections. Is this supported as shown in IMT?
There have been packets drops on CI ports, and wanted to make sure.

I have had time to find out myself. If you can please help?

storm tundra
#

It should be supported. You would be better served by upgrading the A300 to use the available 40G card for the cluster network.

However, even though supported,. this scenario is not ideal due to the severe mis-match between the hosts.

dire sonnet
#

A follow-up. If we upgrade 2 x 10G to 2 x 40G or 2 x 25G on A300. Do we need to upgrade or change RCF file?
Currently we are running ClusterStorageRCF_1.13_shared. Thanks,

storm tundra
#

First the I am pretty sure the a300 doesn’t support 25g

#

Second you shouldn’t need to modify the rcf as you are already have ports ready. You would just connect to a couple of non breakout port

If you do add the 40g,
Add the cards in. Connect to switches. Verify they are up. Add them to the cluster broadcast domain. Then modify the home-port of the cluster lifs. Finally remove the 10g from the broadcast domain and most importantly remove the connections
(There is a bug that seems to pester ONTAP about not being able to send unicast on the ports that aren’t in the cluster broadcast domain)

fringe spoke
#

yup, only 10g or 40g on the A300 for cluster backend

dire sonnet
#

Do you believe with our our configuration: 2x10G on A300 and 1 x 10G on C60 with N9K, would it be the bottleneck of the perofrmance?
AIQ-UM and qos commnds both didn't clearly indicate so. But, indirect accss resulted in slowness not direct access

fringe spoke
#

1x 10G is not supported on the C60. If you want to go 10G there, you have to use all 4 ports of an X60132A card in slot 2

#

you can check hwu.netapp.com on what connections are supported and what hardware is required for each option

glad crystal
#

Isn't is also an issue that the A300 can't be upgraded past 9.16.1 while the C60 is a newer generation which may have gains by upgrading past 9.16.1 in the future?

dire sonnet
dire sonnet
fringe spoke
#

yeah A300 is limited to 9.16.1

#

so if you use this setup only temporary (e.g. during vol move migrations), I think it's fine. Just make sure you minimize cross-HA-pair traffic in the cluster backend to a minimum, and especially try to avoid accessing volumes on the A300 through LIFs on the C60 (as bursts there can easily overwhelm the 10G ports, leading to packet drops and delays or other issues). If you keep the LIFs on the old nodes for example, while migrating the data to the C60, it should be fine

fair mortar