Posts Tagged ‘fex’

FEX configs are retained

Monday, February 15th, 2010

The N5k maintains the config of the FEXs, even when they’re removed (and even though it’s not visible to us). In this article, we’re going to look into this a little further… We’ll start with a working FEX, using Po100 (with Eth1/19-20 being the physical bundle members). (more…)

Some NX-OS features can’t be manually enabled

Wednesday, February 10th, 2010

The whole idea around features in NX-OS has been intriguing to me. It’s a good idea – I like it. It seems very similar to services on many other OSs (particularly *nix systems), however it doesn’t always strike me as having a rich feature set as of now.

For instance, the NX-OS config guides repeatedly refer to the show feature command to look at which features are enabled.  I’ve already mentioned that this command doesn’t exist today in older NX-OS versions – the currently-available NX-OS versions today support this feature (a reader confirmed this on the N7k and I’ve confirmed it on the N5k).  See this article for more info.  There are some interesting behaviors around features, in that some are manually enabled, while others are automatically enabled and disabled as needed.  Let’s dig into this a little deeper for an example… (more…)

Why EtherChannels should be used for FEX interfaces

Monday, February 8th, 2010

I prefer to use port-channel interfaces for the fabric interfaces when connecting fabric extenders (FEXs). If a single interface in the bundle fails, it won’t remove the fabric extender interface – it simply reduces it’s bandwidth. This results in stable, predictable, redundant and resilient behavior. Let’s prove this point. (more…)

Experimenting with Static Pinning

Thursday, February 4th, 2010

What happens when more links are associated with an FEX than are permitted in the max-links parameter? Let’s find out…

(more…)

The FEX information is cached until it’s re-assigned to a different FEX

Tuesday, February 2nd, 2010

Continuing our journey into the world of FEXs on the Nexus 5000 and 2000, today we’re going to look at the behavior of the FEX configs themselves on the Nexus 5000.

All of the configuration and software information (firmware/images) for the FEX (Nexus 2148T) are kept on the parent switch (a Nexus 5000-series switch). Going on this, what happens when the FEX goes offline? Do I lose my config?

A stack of 3750 switches can be provisioned ahead of time, so that as stack members are added (assuming that they’re the correct type/model), no changes must be made – just move on. What kind of behavior is available on the Nexus 5000/2000?

I haven’t found provisioning available yet for FEXs on the N5k, however I did notice that the configurations for the FEX are retained when the FEX goes offline and online. This is great — if the 2148T fails or the fabric interface links are disconnected, the config will still be there (although not visible until the FEX comes back online). This article is looking at it from an association perspective (N2148T association with the N5k, not a configuration (actual port configs on the FEX). With that said, let’s explore this further… (more…)

What happens when the wrong serial number is entered for the FEX?

Sunday, January 31st, 2010

Continuing on the journey of FEXs on the Nexus platform, we’re exploring about serial numbers being statically bound to an FEX instance. It’s possible to allow any N2148T to connect to any FEX instance defined on a N5k. Simply plug in the N2k and off it goes.

This is a good low-maintenance approach, but what if I want to be more granular? What if I have several N2ks and I want to ensure that each N2k is plugged into the correct interfaces on the N5k? The only way to verify this is to tie a serial number to the FEX ID on the N5k, which will only allow that single N2k to come up on that FEX ID… or will it? What happens if the wrong serial number is entered?
(more…)