A Minor Nexus 2K Gotcha, and a Wish

Peter Welcher
Architect, Operations Technical Advisor

Without going into irrelevant details, someone was recently working with Nexus setup in a lab. As you may know, the N7K requires specifying the FEX type. Or did the last time I tried doing without it, a while back. Why, I don’t know, it should be able to figure it out like the N5K does. Perhaps the programmer was squeezed on time. Or needed that extra 5-10 minutes for lunch.

Anyway, they picked N2232TM as it looked like what was on the N2K chassis. The N2K eventually sort of came up, but was showing signs of being stuck in a code download loop. After a while, the person involved (sharp eyes!) spotted the extra -E on the chassis matching the show fex detail output.

Well, with a little help from our friends at Google, when you scrutinize the latest data sheet for N2K’s, you’ll discover the N2232TM is supported on the N7K platform, and the N2232PQ and TM-E models are not. Not yet, anyway, as far as Google reveals. Presumably in upcoming N7K code.

My guess is their VAR apparently helpfully changed the Bill of Materials to include N2232TM-E rather than the non -E version. They probably figured the customer would appreciate the enhanced version. It does have better BER and supports FCoE, which the site does not plan on ever doing.

You might say “should have been paying closer attention to the precise model typecode, etc. And to the supported equipment list.” Well, maybe.The first thing I generally do when looking at a pile of new gear is not check if it is all supported. I kind of assume that went into the ordering process. And I’m not slamming the VAR here, they were trying to do the right thing by the customer.

And now we get to the point of this blog.

When you think about it, why should you or I have to do any of that extra effort (checking if supported)? How hard would it be for the programmer to check the attached model number? The N7K obviously had found it out and displayed it, so the necessary code had already been written! Then, if appropriate, print out an error message to the effect “Configured type does not match actual hardware”, or even better “That FEX type is not supported in this NXOS code release.”

Something similar could be done with SFP/SFP+ issues: echo back what’s detected, and then tell the user it isn’t supported in that NXOS (or IOS) release. Versus the not-very-helpful “SFP not recognized” that I’ve experienced. Leading to trying to recall that obscure show command for the optics on an interface — at least I rarely have had to use it, so it’s not in the short-term memory cache.

That’s why I’m writing this blog. For lack of a little defensive programming in various places, how many people are ending up spending time scratching their heads over stuff like this? And then calling TAC (which you should always do, so that Cisco pays for its mistakes).  If Cisco is serious about making things simple, this sort of thing would be a great place to begin.. Maybe provide better user friendliness about error conditions. Have the CLI tell you what you’re doing wrong, or at least that the code version doesn’t support <whatever it is>. Obviously it can’t do that for a new feature, but it can pretty easily do it for new hardware, even removable hardware.

While I’m doing wishlists (or crabbily venting?), let’s get it all out.

I’m wondering how many readers have attempted bug scrubs. You know, when you scan the bugs in a release you’re considering upgrading to?

Is it me, or are 90% of the bug descriptions that show up in the Bug Navigator completely useless?

Yes, I can see that something cryptic is wrong with “IP multicast, the franistan table in TCAM overflows”. Gee, thanks to the Cisco engineer that put that in the database, that really … doesn’t help. Does it or does it not relate to the PIM problem I’m seeing?  Constructive suggestion: write bug descriptions that make sense to real customers not familiar with IOS or NXOS code internals. Spend a few more characters on communicating.

The third situation I seem to end up in a lot is the Cisco documentation lack of detail dilemma. As in, the documentation doesn’t tell you enough about what a certain command really does, so you have to run off and spend some time doing a test to see what the command actually does. Assuming you want to know all that badly.

Now multiply by the number of people who also have to do the same thing. My favorite of all time is the 6500 documentation that said that “mls qos” quote “enables QoS”. I’d never have guessed. The question I wanted answered was, what is it doing in terms of default trust, queuing behavior, etc.? In some of the small switches, it does dire things in terms of smaller queue depth. So one does need to know!

You can see how often this happens by looking at a lot of the CCIE prep blogs out there. Thanks to the authors for contributing to the general welfare, sharing what they figured out, and saving the rest of us from having to do the experiment.

I also wish to note, no person or company is perfect, and the Cisco documentation is light years better than what I’ve seen from many competitors and technical product vendors. If everything were thoroughly documented, the documentation would probably be 10 times bigger and probably vastly incomplete or delayed while engineers provided the necessary information. I can still wish for more, can’t I?

Maybe we need a term for this sort of minor annoyance or hassle. Is this a geek-nit (or a gnit)? Sort of like a small insect bite. Each one is pretty minor, but when you run into enough of them, they get kind of annoying. If you want to add vendor flavoring, we have C-Gnits and J-Gnits …

We now return you to your regular blog viewing.

Related Links

The latest N2K data sheet as of last week, at http://www.cisco.com/en/US/prod/collateral/switches/ps9441/ps10110/data_sheet_c78-507093.html.

Life Log

I’ve been rather busy lately. Which is generally a Good Thing (sure beats the alternative). Apologies, but blogs come after work, and lately sleep has had higher priority. I had to get this one off my chest, though. I’ve got some happier thoughts queued up. They’re probably coming as small quickly written transmissions.


The vendors for Network Field Day 5 (#NFD5) paid for my travel expenses and perhaps small items, so I wish to disclose that in my blogs now. The vendors in question are: Cisco, Brocade, Juniper, Plexxi, Ruckus, and SolarWinds. I’d like to think that my blogs aren’t influenced by that. Yes, the time spent in presentations and discussion gets me and the other attendees looking at and thinking about the various vendors’ products, marketing spin, and their points of view. I intend to try to remain as objective as possible in my blogs. I’ll concede that cool technology gets my attention!

Stay tuned!

Twitter: @pjwelcher

Leave a Reply