While LACP on VMware ESXi has been around for a while, I noticed that many of our customers do not like to implement this feature in their environment. Reasons vary between, complexity, incorrect understanding on the ‘added value’ and many more. Because of these reasons, I had a discussion on this topic with my colleague and co-blogger Jens. As a result of this, we decided to explain this feature in detail.
Let me first start of going through some of the terminologies used when talking about LACP.
Terminology around LACP
- LA or Link Aggregation is the method of combining multiple physical links in parallel.
- LAG or Link Aggregation Group is the grouping and identification of multiple physical links in a single aggregated link.
- Note: A physical link can only be member of one LAG at the time.
- LACP stands for Link Aggregation Control Protocol. It is the protocol that controls the bundling of multiple links through sending and receiving LACPDUs.
The use of LACP brings many advantages to your setup. We have summarized them for you.
Advantages introduced with LACP
- Increases bandwidth of a single physical link.
- Linearly increased bandwidth
- Increased availability
- Load sharing across multiple links
- Auto configuration
- Rapid configuration
- Deterministic behaviors
To successfully establish LACP between a ESXi host and the network switch, their are a few requirements that should be met. Here are the most important requirements that you should take into account. Many of these requirements are valid for all types of LACP configurations.
- ESXi with LACP can only be established between the host and a single switch or a ‘stacked’ switch.
- The configuration on the switch should be set to static mode ON and the VMware vSwitch should be configured with load balancing method Route based on IP hash.
- Do NOT use beacon probing with IP hash load balancing
- ESXi 5.1, 5.5, 6.0 and 6.5 only support LACP on a distributed virtual switch.
Configuring LACP on vSphere side
Because the configuration of LACP between a ESXi host and a network switch require configurations on both ends, we will explain the configuration in a chronological order which worked for our scenarios. Hence I’m sure that the steps could be performed in other orders, we feel most confident with our method.
The screenshots we took are done in VMware vCenter Server 6.7
First of all, start by connecting to the VMware vCenter UI.
- Creating a distributed switch (right-click on the datacenter object and select Distributed Switch > New Distributed Switch…
- Give the new distributed switch a sensible name.
- Choose a version for your distributed switch (Keep in mind that if you are running an older version of vCenter that LACP has been introduced since 5.1, selecting older dVS versions might to allow you to create a LACP).
- Configure your new dVS with the following parameters
- Number of uplinks
- Network I/O Control
- Creation of a default port group
- (optional) Name of the default port group
- Validate your dVS Settings before submitting
Next, we will create a LACP Group within dVS
Left click on the new dVS > configure > LACP > + New
- Next up, we will add an individual link to the vDS in the LAG
Do NOT migrate VMKernel Adapters or Virtual Machines now, we will first verify if the LAG is accessible and migrate services later.
In part 2 of this blog series, we will continue our configuration on switch side and give you some insight on validation and troubleshooting.