Hi Rene
Say in production network, I have a etherchannel with 4 links between two switched configured, this etherchannel seems to be working fine. But suddenly one of the 4 links failed, and I have to replace the ethernet cable. What is the procedure to replace this defective cable without affecting the live traffic between the two switches. I jest simply change the defective cable, and the Etherchannel will now use all 4 links. What happens in this scenario.
Thanks
Palani
Palani,
In most cases you can just simply replace the cable without worrying about it. If you wanted to take extra precautions, a fool-proof method would be to issue a âshutdownâ on each affected interface, replace the cable, then issue a âno shutdown.â This one should be really easy, but be sure to make it appear difficult so you are more appreciated at work 
Hi,
I am also trying to get my head around this as I just missed this question on a boson practice exam. It is similar to what you have setup in your lesson except they use a router instead of a server basically.
From what you are saying it seems like if you have multiple devices connected to a switch that will be using the Ether link that using the src will not bind it into a contract with a certain link but allow it to use all of them.
Where as if you use dst it will bind to one of the links and continuously use that thus not using the other links and there would be no load balancing effect.
What is confusing me is if you use src-ip and you had four devices connected to switch A does the port-channel load-balance part of the command try and make it use different links to try and load balance, and then the src-ip is the key that allows it in this case because of multiple devices? and since there is multiple devices four in this case it will use the four different links to send.
and here is the specifics of what I am unsure of: So if we setup dst-ip on the switch connected to the server it will use the same path for each of the different PCs but since those PC connections from Switch A side all use different links to get to Switch B and the server (because of the port-channel load-balance src-ip command) it has the effect of making the server return information (using the dst-ip) achieve load balancing.
meaning the real key here was not the server side but the effect of multiple pcs on the PC side that used src-ip. is that the correct thinking?
if thatâs is the case then thatâs making sense now that I put the thoughts in written form.
Using these forums really help me when I have to try and formulate my thoughts into written language and post on here, not to mention the reinforcement and feed back by others. Great learning tool that helps cement things for me. love emâŚlove em!!!
Hello Brian.
If I understood your explanation correctly, then yes you are correct.
Essentially if you have multiple devices communicating with one device over an etherchannel link, then traffic that goes from the multiple devices towards the single device should be load balanced based on source IP. There are many source IPs so each one will use a different physical link and load balancing will be achieved.
Traffic from the single device to the multiple devices should be load balanced based on destination IP. Again, there are many destination IPs so each one will use a different physical link and load balancing will be achieved.
I hope this has been helpful.
Laz
Thanks!
I got to thinking after I wrote this post. I was thinking now its pretty easy to know what to do if I have multiple devices on one switch and a single device on another switch like a server or router.
However what about if you had say multiple devices on each switch. For example Switch A has four PC connected and then Switch B has four PC connected and Switch A and B are connected by a 4 port etherchannel?
I know you would sit one side or the other side up as src-IP/mac but after setting that up would you set the other side up also as src-ip/mac or just set it up as dst-ip/mac or would it even matter one way or another when setting up the second device?
would it use a random port on the way back or would it be forced to use the same port that it received the information on when returning.
If it was forced to use the same port to return then it does not matter what you choose on the other device src or dst because its going to do what its going to do.
Its more of a curriousity question and how to answer a question where there is multiple devices on both sides.
I guess normally traffic is going somewhere and not being sent out randomly which is why we have IP addresses and Mac Addresses even UDP traffic while not controlled is still going somewhere specifically.
but if these are computers that might be talking to different computers at different times then that scenario could happen and if thatâs the case I would think you would set both sides to src-ip/mac and whoever setup the initial data when the other device on the other switch responded it would use that same port to speak back.
I think I read something about once a connection between devices Is used with etherchannel the port is normally used for that connection so that load balancing is not truly load balancing.
I donât want to get to much in the weeds here on this question just wonder how I would answer a question like that with multiple devices on both sides!
My guess is that you sit both sides up as src-ip/mac however the connection is going to use a single port for all the communication between whatever two devices happen to be communicating at the time. So in other words it will act like a src-ip/mac to a dst-ip/mac. However it needs to be setup as src on both sides because of the initiation of communication from multiple devices.
I woke up this morning started studying and had another similar practice exam question from boson on ether-channel I am going to stare at this a bit this morning as I try to wake up as it has some good information as well see below:
Hello Brian
In order to determine the best load balancing technique, you must determine the nature of the traffic. The multiple devices on one end and a single device on the other end is a special case that must be addressed. In most cases, if you have multiple devices on both ends and the connectivity between the two âendsâ is for the most part random, then no special configurations should be implemented. The default is to use the source and destination IP to determine the port (using a hashing algorithm) which is sufficient for such situations. For more information on load balancing and redundancy, take a look at this Cisco documentation.
I hope this has been helpful!
Laz
Hi Rene,
Regarding Etherchannel Load Balance âŚ
What is the meaning of âSrc XOR Dst IP Addrâ
Is it, If any IP Change either src or dst the flow will be different âŚLike
SRC(1.1.1.1) --------DST(2.2.2.2) Flow-1
SRC(1.1.1.2)--------DST(2.2.2.2) Flow-2
SRC(1.1.1.2)--------DST(3.3.3.3) Flow-3
br//zaman
Hello Mohammad
The command src-dst-ip provides the load distribution method of source XOR destination IP address. Essentially this means that each source and destination pair are taken into account in the hash algorithm. Your example is essentially correct.
I hope this has been helpful!
Laz
We create EtherChannel to maximize the throughput. So if we have 8 100M links we can use them as a single 800M link using EtherChannel. While using source-destination MAC load balancing we restrict the source port to destination port connection.
In this case, can it utilize the 800M BW? What I am trying to understand is, does the ether channel use one physical link but speed of the entire bundle?
Hello rosna
This is an excellent question and is vital to understanding how EtherChannel works. I will use the example topology you described in your question.
If you are sending a file for example over your 8 port etherchannel bundle using FTP, you will never see speeds reaching 800Mbps, but only 100Mbps. This is because each session, or stream of data can only use one physical link. In other words, if you use the same addresses and session information, the data sent will always be over the same port in the channel. This method prevents out-of-order packet delivery.
The advantage of EtherChannel is not the speed it provides to one individual session, but the ability to load balance many different sessions from and to several end devices over the various available channels in the EtherChannel providing more bandwidth overall as well as redundancy.
Distribution algorithms can be chosen based on source or destination IP or MAC address as well as source or destination port number on the switches. So, if you have 10 PCs on one end communicating with another 10 PCs on the other, the source and destination MAC and IP addresses will be used to distribute the sessions across all links thus taking advantage of the total bandwidth available.
You can find additional information about the available distribution algorithms on Cisco devices at this Cisco Documentation.
I hope this has been helpful!
Laz
Isnât it counter-productive to put a trunk over an etherchannel? I think i know why it isnât, but could you please explain?
also when is it better to use PAgP vs LACP and vice versa?
Thank you and sorry if these questions were already asked, i couldnt read through all 65 replies.
Hey Austin,
Having more bandwidth over a trunk is a great thing! We will be carrying traffic for many subnets/vlans over a trunk so the bigger we can make the pipe with an etherchannel the better.
Also LACP is a protocol that can be used by many different computers and vendors you can run LACP on servers and switches which is great. PAgP is a Cisco proprietary protocol so it can only be ran between Cisco equipment. I hope this helps!
Thanks,
Scott
I understand about the trunk. My understanding was a bit flawed. I was thinking we were combining a path and then seperating it again.
Also, if we are on a network built entirely with cisco gear, is there a better choice between LACP and PAgP? If so, why and when is one better?
Thanks in advance!
I dont think there is a difference between the two in regards to pros and cons. The only advantage I could see is that PAgP might be âquickerâ to configure since the default protocol used for etherchannel is PAgP on Cisco equipment. After some googling there was no clear advantage I seen. One advantage of both LACP and PAgP is they help prevent loops manual etherchannel configurations do not allow for this. I hope this was helpful!
One advantage I see that LACP had over PAgP is that it supports stacked switches.
I would just stick to LACP, even if you only have Cisco hardware. You might want to use LACP not only between switches but also between switches and servers. Itâs easier to use LACP everywhere and since itâs an industry standard, you canât go wrong with it.
I can explain in detail the parameters of the output, for example the load and also the other fields
Hello Jesus
I hope this has been helpful!
Laz
Please I want to know why we call PAgP and LACP negotiation protocol?
Actually what is a negotiation protocol?
Thank You
