LAG

[cable guy]

LAG configuration for Link Aggregation - combining radio link A and B to achieve a near double throughput end to end

I have 2 WS-8-150 switches on the bench, both in the same subnet 1.41 and 1.42
LAG is enabled on port 2 and 3 on each switch
I have 2 radio links in between, port 2 to 2 and port 3 to 3
one running 35 meg and the other running 45 meg because that is my real world scenario (long distance)
I want to achieve a LAG throughput of 80meg~ from point A to point B
ports 2 and 3 are set to LACP-A
STP has STP on LAG enabled and tried with it disabled

I have 2 PC's running iPerf running a speed test from switch port 1 to switch port 1

Instead of both links running traffic, only 1 does, the one that STP chose as the root, the other is the alternate
shutting off STP results in nothing working

Will LAG provide some sort of balanced traffic flow across the 2 links ?

what am I missing to make it do this ?

Dave

[sirhc]

This is not how LACP or Static LAGs work.
LACP and Static LAGs are an industry standard.

LACP = IEEE 802.3ad

https://en.wikipedia.org/wiki/Link_aggregation

You can not take two links such as 20 Mbps each and with LACP or Static LAG and make a link that will pass a single 40 Mbps stream. What you end up with is a link that can pass upto 40 Mbps in total aggregate data throughput via multiple streams but not a single data stream greater than 20 Mbps.

Everything in our switch uses industry standard protocols such as STP, RSTP, LACP, VLANs, and so on.

You can search the web and read about all these protocols.

[cable guy]

Yes I understand all of that clearly, but I do not know what specific settings are needed in your switch, no help pop up or file

I have spent hours trying every combination possible

I need to understand how to configure the 8-150 to make LACP work
Are the 2 ports on one side set for LACP-A and the far end set for LACP-P, or do both ends get set for LACP-A
I presume both paths have the same Key / SID number, such as 1

Do I set LAG to use Source/Destination Port

Is STP on LAG supposed to be enabled

Is STP supposed to be enabled or disabled

Dave

[sirhc]

When I was first working with Eric on this feature of the switch firmware I set up 2 Netonix switches in a LAB and played with them while reading the WIKI information and got them working.
https://en.wikipedia.org/wiki/Link_aggregation

5. LACP mode:
active: Enables LACP unconditionally.
passive: Enables LACP only when an LACP device is detected. (This is the default state)

This section defines the Role:
LACP-A = LACP mode active
LACP-P = LACP mode passive
Static = For Static LAGs

LACP Advantages over static configuration
Failover occurs automatically: When a link fails and there is (for example) a media converter between the devices, a peer system will not perceive any connectivity problems. With static link aggregation, the peer would continue sending traffic down the link causing the connection to fail.

Dynamic configuration: The device can confirm that the configuration at the other end can handle link aggregation. With Static link aggregation, a cabling or configuration mistake could go undetected and cause undesirable network behavior.

So I assume you want LACP since your going from switch to switch over Static LAG since LACP can drop a link and operate normally whereas with a Static LAG if 1 link from a group drops things do not adjust automatically and data will be lost.

You might as well use LACP-A (active) on both sides, you could have one side Active and one side Passive but I use Active on both sides.

The Key has to be the same on all ports of the LAG on both sides (a unique LACP Key for this Layer 2 segment)

Timeout is just how fast the LAG will drop and reacquire the link in the LAG, I use Fast.

As far as Priority just use the default value of 32766 unless you do some more reading on LACP and have a reason to change it.

You configure the LACP system priority on each switch running LACP. LACP uses the system priority with the switch MAC address to form the system ID and also during negotiation with other switches.
The system priority value must be a number in the range of 1 through 65535. The higher the number, the lower the priority. The default priority is 32768.

Lastly the Aggregation Mode at the top simply specifies how the LAG will distribute/balance the outbound data going across the LAG.

Destination MAC | Source MAC | Source/Destination IP | Source/Destination Port

Once again since our switches follow industry standard protocols anything you want to learn about those protocols can easily be found using Google.

For instance to learn more about LACP Priority simply Google "LACP Priority" and do some reading.
https://www.google.com/search?q=LACP+pr ... e&ie=UTF-8

[jakematic]

Everything Chris said is spot on, and I'm a visual kinda guy who likes to see examples so maybe this will help.

This switch is using SFP but it is the same with copper, just make the settings the same on both sides.

You'll never fully balance the links.
Depending on the type of traffic (e.g. many clients to many destinations) you can experiment with the aggregation mode to see what works best for you.

[cable guy]

Thank you for the screen shots, the are always preferred and appreciated
if I have an example of a working copy then I can quickly copy and use it

Have a look at my screen shots, there is something wrong
I'm seeing not active on both links
and root on one and alternate on the other
does not match what I see on yours and traffic will only flow on one link or the other
if I fail the root link then the other link becomes root and traffic will flow on it
traffic does not flow across both
i ran iperf with a far higher throughput than the radio links can pass and traffic only flows down one link
until I fail that link and then traffic will flow the other
but not over both as all of the documentation I have read says it's supposed to
here are the screen shots

Dave

[Julian]

single streams don't get forked and load-balanced across a lag and reassembled at far end, don't forget. I'm pretty sure iperf is a single tcp stream by default, do you have any information on your iperf configuration?

[cable guy]

You're correct, the documentation said that streams would not be split up.......my bad, lack of experience
a simple hash would direct streams down either path and not break them up for fear of one segment arriving before the other
therefore Lag could never balance evenly

I'll switch to jperf and select multiple simultaneous tests

I am still mystified why I'm seeing not active on both paths on one side and root and alternate on the other side

Dave

[jakematic]

I am still mystified why I'm seeing not active on both paths on one side and root and alternate on the other side

Dave

Compare that the TX/RX columns to my screenshot - you aren't getting LACP to negotiate for some reason.

What I would do:
- make sure that exclamation on the device tab is cleared on that one switch
- break down the LAG on both sides
- let the links stabilise
- create the LAG with 2 ports on side B
- create the LAG with 1 port on side A let it stabilise
- add the second port to the LAG

Be sure to click save/apply after each change

[agidi]

Hello Netonix Gurus.. newbie here

I was giving this LAG thing a shot on a live link

We have two towers with WS26400AC on each one.

Linked with an airfiber, which was maxed out.
So we bought and installed a new link with some airfibers HD

Instead of taking down the old link i thought about using LAG.

Following the previous suggestions

On the remote site, i added a key and enabled LAG for the ports, with only Source/Destination IP checked as per the examples.
As soon as I save the remote site, I lose access to it, and for some reason wont go back.

We determined this after doing the local and remote at the same time and loosing both. LOL


What could cause me to loose access to the remote side on an improper LAG config?

thanks

EDIT.... an improperly tagged port.. I´ll test next week and report. thanks