networking connections on A800

[regal helm]

We are going to headswap A400 to A800. There are only two X1146A cards for networking upon the quote. If you can please help me out on questions below:

1. The X1146A should be for 100G client connections(we have 100G backbone switch). Correct?
2. Since there are total of two X1146A, so, one card for each node. But, shouldn't we have two such cards on each node for redundancy?
3. Since there are no any other networking cards listed in the quote, then e0a and e0a only onboard port(Upone HWU, there is even no e1a port available) can be for cluster switch ports, Correct?

[subtle lagoon]

hi. this picture from HWU should answer your question 3. As for question 2, that depends. Technically a second client-facing card would increase redundancy but in practice it doesn't really matter since LIFs can freely migrate to the other node if required. And to answer question one, if there are only 2 X1146 cards for both nodes, then you will not have any client connectivity unless you move cards from the old controllers to the new ones. For Head Upgrades, you usually have to specify additional network cards that you require, otherwise you will only get the bare minimum

or, to put it more directly: no, you need that X1146A card for cluster/ha connectivity and it has to go in slot 1. you cannot do client networking through it

[regal helm]

Then we only have one X1146A for the client networks in entire 2 nodes HA, and only e0a for the cluster connections.

As for your comment on my third question, on current A400, there are X1151A's and 1161A's for respective cluster and client connections. But, they cannot be used on A800 upon my understanding to HWU.

So, the quote is not good. Agree?

[subtle lagoon]

Then we only have one X1146A for the client networks in entire 2 nodes HA
Again, maybe I didn't understand you correctly, but if you only have TWO X1146A cards for TWO nodes, then you have no client connectivity since you need those cards for Cluster/HA (1 in each node, in slot 1)

[regal helm]

Yes, you are correct. There will be no any cards for client connections.

[subtle lagoon]

and yes, the X1151 is only supported in the A400, and the 1161 is an onboard mezzanine that cannot be used anywhere else

so yeah, I would complain to the engineer who did that config/quote

[regal helm]

Yes, Thank you!

[deep flume]

Hi David.
Without analyzing the quote, and knowing more about:

What disk elements you are bringing over and your migration strategy, such as having a NS224 shelf from the A400 (which doesn't have internal drive bays), requiring an X1148 to swing the shelf over, since you can't move the drives to the internal bays of an A800 without a wipe.

The other thing you should be aware of is mixing self-encrypting and non-encrypting drives.

The base configuration of an A800 includes 1x X1146 card per node in slot-1, giving you 2x 100Gb cluster uplinks per node. The onboard ports e0a/e0b are used for HA, also 2x 100Gb. You would then need to addon 1x more X1146 card per node for 2x 100Gb network uplink. I don’t know how your quote is written, but adding 2x X1146 cards for a 2-node A800 sounds normal.

I would recommend that you reach out to your vendor, and/or NetApp sales if you have any questions or concerns.

Lastly, hope you are getting a screaming deal, the A800 may go end of sale next year as it was released in 2018; the AX0 line is replacing it. The VER2 hardware was released earlier in the year, but that was to get around chip shortages. I assume you have a A400 VER1 and potential end of sale is what is motivating the head swap, or your support is running low and it can be cheaper to support a new controller purchase with minimal drives and then move over existing drives, inheriting that support. Once a controller goes end of sale, typically it receives 5-years of hardware support after that.

[subtle lagoon]

mixing self-encrypting and non-encrypting drives.
this is only a concern for NSE drives (FIPS compliance). If you have non-FIPS self encrypting drives you can freely mix and match them with non SED drives

Other than that we basically established all that alreaday but thanks for summarizing it 👍 I guess the partner/sales rep has just done a bad job here (or rather no job at all 😄 )

[regal helm]

What you said sound correct.
According to HWU, the onboard port e0a is used for the cluster connection to the cluster switch ports, and e0b for HA. So, two 1146 cards in the quote, one for each node in slot 2/3/5, should be for client connections.
Thus, now the config/quote should be fine. Is my understading correct?

[chrome sail]

e0a to e0a and e1a to e1a are used for the cluster network connection.
e0b to e0b and e1b to e1b are used for the HA connections

Anything beyond that is for customer networking

[regal helm]

According to HWU, there are no onboard e1a, or e1b, only e0a, and e0b. Correct?

[chrome sail]

Since there is a card in “slot” 1 it is not “onboard”. Hwu will show exactly one card for use in slot 1, x1146 which is for metro cluster/ HA. This is a mandatory card.

Look at the “supported connections” on hwu. That should answer your questions

[subtle lagoon]

the "0" in "e0a" stands for "onboard" (i.e. "slot 0" which is the mainboard). the "1" in e1a stands for "PCI slot 1", so it is not an onboard port. It is still a port that has to be installed in every system for it to work

"onboard" doesn't mean "required" (and vice-versa)

[regal helm]

OK. The base A800 does include one 1146 already, so, plus additional 1146 in the quote, we will have two 1146 on each node, total of four 1146 for 2 nodes HA

[subtle lagoon]

"required" and "included" are two different things. AFAIK (it's been a while since I last had to look at quotes so TMAC might correct me here) the card is not Automatically included in the base system, it is an extra part that is quoted separetely (and listed on the quote)

[chrome sail]

That’s actually one thing that I despise on Netapp quotes!
Like the fas9000 you automatically would get the sas card even though it’s not listed but they list the network card. I mean if it’s mandatory why include the line item at all???

Id have to build a quote and see what it does today to tell you the truth

[regal helm]

I am sorry, I am confused. Since our quote only include one 1146 for each node, so, are we getting one 1146 or two 1146 for each node?

[subtle lagoon]

@regal helm I would just contact the person who sent you the quote and make sure that they know what you expect to be delivered and have them cross-check that everything is there. Probably even in writing (mail) just to be on the safe side. This makes it easier for you if they later claim that you ordered it wrong

[subtle lagoon]

I think we went over this already yesterday. The card you got is the cluster card that is required and not available for client networking. If the quote say 2 cards then it's 2 cards, not 4. At least from my experience this is how it works here in Germany

But to be sure I would confirm with the partner (if you're buying from a partner) or NetApp sales rep (if you're buying through NetApp directly)

this is the only reliable path to take to be 100% on the safe side

It is much better to find out that they need another two cards in the quote before you buy it, since if you buy the cards later they will be more expensive usually (the reason being different rebate structures for standalone parts than for complete systems/projects)

[regal helm]

Yeah, I will definitely check with him, but just to do some homework before then…

[subtle lagoon]

these cards are like 1000 bucks or so. So when in doubt just have them add 2 additional cards anyway which shouldn't increase the whole project cost by much, but you'll be on the safe side

[chrome sail]

Ah-hem. Aren’t They like $10k list!??

I know the x1148 certainly is.

[subtle lagoon]

yeah, I meant price within a project which is usually a fraction of that

nobody pays list prices anyway 🙂