#X60132A-C and X60130A-C are used for Host Connections or Cisco switches?

1 messages · Page 1 of 1 (latest)

frigid crane
#

We are going to implement AFF C60. These two NIC cards in the quote. Can you please tell me which card should be for HOST connections, and which for Cisco intercluster switch?
Cisco Intercluster Switches: N9K-C9336C-FX2>

X60130A-C IO Module,2PT,100GbE,-C
X60132A-C IO Module,4PT,10/25GbE,-C

If either one can be used for Host or Intercluster, what is your recommendation.

tough sparrow
#

The X60130 is for the cluster network and HA. The X60132 is for host connectivity and intercluster. HWU has all the details

frigid crane
#

Can X60130 be used for Host?
I cannot understand HWU. Thank you!

tough sparrow
#

yes X60130 can also be used for host but then you need another one for HA/Cluster

#

since the X60132 cannot be used for cluster/HA

frigid crane
#

What is this X65404-N-C? Can it be used for HA/Cluster?

If not, we are then missing one for HA/Cluster?

tough sparrow
#

x65404 is not a network card, no

frigid crane
#

Can N9K-C9336C-FX2 support 25/10G?

tough sparrow
#

yes, with an adapter or breakout cable

frigid crane
#

I only see in the quote, so, if the first one is for Host, and 2nd one for Intercluster, then we have no one for HA/Cluster.

tough sparrow
#

you could use the 60130A for HA/cluster and the 60132A for host and intercluster?

obtuse frost
#

The C60 uses two cards for cluster/ha. Pretty sure the A ports go to the switches (or use point to point). Lower end platforms allow for both to be used

#

The 25g cards are used for all other data connectivity

#

Single 100 card in slot 4, port a to switch 1 and port b to switch 2.
If you have 2 100g cards 2a to switch a and 4a to switch b
Looks like b is allowed on this platform for data.
25g is also allowed. You will need breakout cables.

#

As @tough sparrow mentioned, consult hwu. Choose the c series. Select c60 and an ONTAP release then look at supported configurations

tough sparrow
#

I feel like we had this same discussion before, only with an A800 and the X1146 and X1148 cards 🤔

obtuse frost
#

The 800-> always uses the 1148 for ha/cluster. the onboard ports are based on the Mellanox card, the same for the automatic card in slot 1

#

It also always uses the 1148 for storage. The 1146 was added for customers that must use optical connections to switches. Optics are not supported in the 800 unless in the 1146 at which point you cannot do any rdma/roce connections

tough sparrow
#

yeah, I meant we had this discussion on this Discord here before where someone wanted to use the X1146 as cluster card instead of the 1148 or something. But maybe I'm misremembering things

obtuse frost
#

I also corrected my above post and removed the comment

tough sparrow
#

it just felt like a strange déjà-vu effect ^^

frigid crane
#

We are going to use X60130 for Host, one per node., because we want to use 100G for host.
X60132 is 4 ports card, you meant 2 ports for Intercluster switch, the other 2 for HA/Cluster?

obtuse frost
#

Please refer to hwu for SUPPORTED configurations

The cluster/ha use the same ports cards need to be in certain slots to be supported. Hwu is not that difficult to use to figure this out

tough sparrow
obtuse frost
#

It can. It’s in hwu as supported SWITCHED configuration only. although the footnote indicates switchless is ok also. X60132 must be in slot one or two

obtuse frost
tough sparrow
#

ah, I stand corrected. Nice find. I am still confused that it doesn't show up in the "supported cards" as Cluster/HA card then 🤔

frigid crane
#

Can X60132 --4 Ports be used for InterCluster switches (2 ports) and Cluster/HA (2 ports)?

If not, then the quote didn't include NIC for Cluster/HA.

tough sparrow
#

both configs above (here is the second one) require 4 ports for cluster, so you cannot split them 2 cluster 2 frontend, no

#

even with that config I think you would still need the 100g card connections for HA and the 4x25g are only used for the cluster network (as HWU says "Cluster" and not "Cluster/HA" there) but I haven't found the document that confirms that

frigid crane
#

Should usually NIC for Cluster/HA be in the quote?
Is it possible by default the NIC should be included wihtout the need to be shown in the quote?

obtuse frost
#

The card that is slot 2 is the default card for cluster/ha.

#

So the basic requirements with new hardware is that ONTAP will use RoCE nics. Both cards are RoCE. So they should pass the correct info. With the card in slot 2, all 4 ports are used (a/b to switch 1 and c/d to switch 2)

#

Pretty sure as long as the 100g card is not in slot 2 it can be used for data. The question is why wasn’t the correct networking connection asked for?

frigid crane
#

One more follow-up.
C60 requires ONTAP 9.16, upon my understanding.
If all the other existing nodes are 9.14, do we need to upgrade them all to 9.16 first before we can add C60?

tough sparrow
obtuse frost
#

If you do the mixed mode, you have a couple options later on.

Right now, I would personally stand up the new units, boot to the special boot menu and upgrade ONTAP to the most current patch version. (Option 7). Then you will need to let them reboot to the license screen and setup the node addresses.
From the main cluster, there is an option to “cluster add-node” for allowing mixed versions. You will need to specify it.

Also depending on his so you want to upgrade to the cluster, you may wish to choose to use the cli to update only the old nodes. If I’m not mistaken, the GUI will update all nodes even if the same version

#

Alternatively, you can just boot the new controller to see what code is there and update the cluster to the same version first, then add the nodes.

Plenty of ways to make this work

frigid crane
#

We may want to speed up the process by adding C60 first, then do ONTAP ugrade later.
Are there any concerns to run a mixed version, and upgrade later?

tough sparrow
#

it's not a very well-tested scenario so I would strongly suggest to upgrade ASAP after the node join. I wouldn't wait the max. supported timeframe of 90 days

#

and the upgrade doesn't take long (maybe an hour or so), so there's not much speeding up by waiting anyways

oak crest
#

I didn't dig very much, but Option 7 only supports http(s) download sources, I'm guessing. HTTPS would be a problem if it doesn't ignore certificate validation. I just used scp to the systemshell after unlocking the diag account and setting the passwords for admin and diag and then adding the package to the registry and just ran an "old school" manual upgrade

tough sparrow
#

python -m http.server 8080 would still have been easier I guess, but good that you found a solution that worked