VMware Cloud Community
trent9645
Contributor
Contributor

Link Aggregation - Alcatel Lucent

Need help configuring Link aggregation between an ESXI 6.0 host and an Alcatel- Lucent OS6450-P48 switch.

What I have done.

ESXI Host:

Vswitch0

Load balancing: Route Based on IP hash

Network Failover Detection: Link status only

Notify Switches: Yes

Failback: Yes

Active Adapters: vmnic0-3

Alcatel-Lucent Switch:

I have tried both Static and Dynamic

Static:

static linkagg 2 size 4

static agg 1/23 agg num 2

static agg 1/24 agg num 2

static agg 1/25 agg num 2

static agg 1/26 agg num 2

vlan 55 port default 2

Dynamic:

lacp linkagg 2 size 4 actor admin key 5

lacp agg 1/23 actor admin key 5

lacp agg 1/24 actor admin key 5

lacp agg 1/25 actor admin key 5

lacp agg 1/26 actor admin key 5

vlan 55 port default 2

I have also tried configuring the link aggregation group "2" to trunk vlan 55 but it didn't change anything.

Currently, if I just have my regular config on my switch and plug all 4 wires in then I can still get to the ESXI host but I don't think it is doing any load balancing over the 4 cables. Once I configure static or dynamic link aggregation on the physical Alcatel- Lucent switch I can no longer get to the ESXI host unless I unplug all but 1 cable.

Reply
0 Kudos
14 Replies
daphnissov
Immortal
Immortal

What is your goal or requirement when using a LAG with an ESXi host?

Reply
0 Kudos
trent9645
Contributor
Contributor

Just to share the load between all the virtual servers between the 4 lines.

Reply
0 Kudos
daphnissov
Immortal
Immortal

Are you aware that you do not need to use a LAG to achieve this goal? The default algorithm of "route based on virtual port ID" will achieve this naturally just by having multiple single uplinks for a given vSS. This is a far simpler configuration and does not require any specialized configuration on your upstream L2/3 devices.

Reply
0 Kudos
trent9645
Contributor
Contributor

My understanding is that "route based on virtual port ID" can only send data through one uplink at a time. And "Route based by IP hash' can send data through multiple uplinks at a time.

Reply
0 Kudos
trent9645
Contributor
Contributor

Here is an article I found that explains the differences between the options on the ESXI host

Understanding NIC Teaming and Load Balancing Policies in virtual switch – Virtual Reality

Reply
0 Kudos
daphnissov
Immortal
Immortal

Yes, a virtual NIC can only send traffic through one physical uplink at a time, but all VMs are spread across all links. By the way, even with a LAG and route based on IP hash *still* may not achieve a load balancing of NICs as a session between two nodes is still limited to the throughput of only a single NIC. It's a common misconception that data can be "sprayed" evenly amongst all NICs in a bond. So route based on originating virtual port ID still achieves a load sharing but across all VMs in the vSS.

Reply
0 Kudos
trent9645
Contributor
Contributor

Right, 1 computer to the server is limited to 1 NIC. However, if I set Route based IP hash then say 100 clients could be using line 1 and 100 different clients could be using line 2. Correct?

Reply
0 Kudos
daphnissov
Immortal
Immortal

It's a possibility, yes, and a LAG would help here if you have multiple concurrent connections to different hosts. But it's important to point out if Server A is talking to Clients A, B, and C, each of those sessions *may* be on different physical links, but the session to each client is still limited to the physical limitations of the uplink selected. Server A does not evenly distribute traffic to Client A across multiple uplinks.

Reply
0 Kudos
trent9645
Contributor
Contributor

Yes, that is why I want LAG setup and working on Route based IP hash because we are a school district so there are several different hosts connecting to the same server at any given time.

Reply
0 Kudos
trent9645
Contributor
Contributor

Looks like it needs to be a  "Distributed switch" the esxi host, I have it set up as a standard switch; however, I don't have the option to create a "Distributed switch" I think it is because we are running the free version of esxi? I believe we need to buy this below to be able to create a distributed switch for this matter and to be able to connect our 2 ESXI host and replicate the virtual servers between the 2 physical servers. At least that is what I am gathering, does that seem right to you?

https://store.vmware.com/store/vmware/en_US/pd/productID.285644500

Reply
0 Kudos
daphnissov
Immortal
Immortal

If you're using ESXi free you can use static LAGs. If you have ESXi with an enterprise plus license (or some other license) which entitles you to use a vDS, you can use LACP but this requires vCenter Server as well.

Reply
0 Kudos
trent9645
Contributor
Contributor

How do I configure the Static LAG on the ESXI host?

Reply
0 Kudos
daphnissov
Immortal
Immortal

https://kb.vmware.com/s/article/1001938

You just put the physical switch in static mode, then change the vSS's uplinks to Route Based on IP Hash. If your management vmkernel port is sharing this same vSS, you must change the physical switch first before making the port group change.

Reply
0 Kudos
trent9645
Contributor
Contributor

Believe I have tried that with no luck...

Reply
0 Kudos