PDA

View Full Version : Problems with Link Aggregation and a Mini




YinZhouWang
Dec 29, 2012, 01:28 AM
Hi there,

Santa brought me a huge pile of useless crap, but also new network equipment: a Cisco 20-port managed switch (SRW2016-K9)

I'm currently trying to set up a new network environment. For better performance I want the designated server (a Mac Mini 2011 running 10.8.2) to use link aggregation on two ports (the native ethernet port and one via a Thunderbolt adapter).
Both connections receive their IP addresses through DCHP from another router when plugged in > connection established.

I then created a network bond in sysprefs/network ('manage virtual interfaces' > 'new link aggregate' with both ports. No problems so far.

On the switch I set port speed to 1000Mbit and full duplex mode for both ports (Port Management/Port Settings). Then I added both ports to one LAG enabling LACP.
However, only one member - the Thunderbolt Ethernet - was recognized and went green (OS X/sysprefs/advanced). The light for the native port is blinking, changing between red and orange. The error messages 'Invalid Link' ('The link state was not valid, i.e. down, half-duplex, or wrong speed') and 'No Partner' ('The port on the switch that the device is connected to doesn't seem to have 802.3add Link Aggregation enabled.') are changing accordingly.
The port on the switch in which the native ethernet is plugged into is 'down' (no LED).
I checked the physical connections, exchanged cables and used different ports - no avail. When I connect two Macs (one of which is the above mentioned Mini) with two patch cables directly, the connection works just fine.
I further tested link aggregation in a similar setup (Mini 2012 and Display Ethernet and a Retina MBP with two Thunderbolt Ethernet adapters). > no problems AT ALL!

Any ideas?

Any hints are greatly appreciated!



LeicaMan111
Dec 31, 2012, 12:14 PM
Do you have another switch you can try it with?

assembled
Jan 1, 2013, 04:44 PM
The Apple LACP implementation is not quite as robust as it should be.

For a home network however, I would question its usefulness...

freejazz-man
Jan 8, 2013, 03:18 PM
try not manually configuring the ports and letting them auto-negotiate

I would configure the LAG on the switch first, then add the connections in osx. this shouldn't make a difference but your setup should be working anyway so let's not let logic get in the way of things

you sure it wasn't in a spanning-tree wait state when you first plugged the ethernet in? try and put the port in portfast so it can negotiate immediately?

Umac-de
Jan 9, 2013, 07:22 AM
In either case - you know that you will not get a 2 Gbit connection between 2 devices?
LACP only can user/device based load balancing... :cool:

jasonvp
Feb 1, 2013, 09:56 AM
On the switch I set port speed to 1000Mbit and full duplex mode for both ports (Port Management/Port Settings). Then I added both ports to one LAG enabling LACP.

With Gigabit Ethernet, you want to enable auto-negotiation, and let the switch and Mac figure it out from there. So go back and do that, first.

I'm at work right now but my home network is set up similar to what you want to do; I'm using a Cisco 200-series (no CLI like your 300-series, but otherwise identical) and a Mac Pro. I have it working with an 802.3ad bundle between the two, along with .1q trunking over said bundle. What I can't remember is whether I enabled LACP on the bundle or not. I'll check when I get home and reply back here. Either way, first thing's first: get auto-neg running on both ends of both links.

jas

jasonvp
Feb 2, 2013, 12:19 AM
What I can't remember is whether I enabled LACP on the bundle or not.

LACP is enabled. So you should be good if you just reset your ports on both ends of the links to autonegotiate.

jas

mvmanolov
Feb 5, 2014, 12:24 PM
In either case - you know that you will not get a 2 Gbit connection between 2 devices?
LACP only can user/device based load balancing... :cool:

Can someone please explain to me why this is the case?

brand
Feb 9, 2014, 01:44 PM
Can someone please explain to me why this is the case?

Because that is how it is designed. If you want speeds faster then 1Gb for a single network computer then you need to step up to 10Gb. LAGG/LACP are not the solution for obtaining faster than 1Gb speeds for a single network computer.