---
title: "Connect a load balancer"
description: "Put a TIR Auto Scaling Group behind a Web traffic load balancer so requests reach every VM of the group, and detach it again from either side."
content_type: "how-to"
product: "tir-auto-scaling"
platform: "TIR"
url: "/tir/tir-auto-scaling/how-to/connect-a-load-balancer/"
topic: ["network"]
---

# Connect a load balancer

A load balancer gives your Auto Scaling Group one address and shares the requests among its VMs.
[TIR Load Balancer](/tir/tir-load-balancer/) covers the load balancer itself. To make the same
connection from the load balancer's side, see [Connect an Auto Scaling
Group](/tir/tir-load-balancer/how-to/connect-an-auto-scaling-group/).

## Before you begin

- An Auto Scaling Group in your project. See [Create an Auto Scaling
  Group](/tir/tir-auto-scaling/how-to/create-an-auto-scaling-group/).
- A load balancer of the type Web traffic (HTTP/HTTPS) that is **Running** or **Stopped**. You
  can make one from the group; see [Make a load balancer for the
  group](#make-a-load-balancer-for-the-group). To make one separately, see
  [Create a Load Balancer](/tir/tir-load-balancer/how-to/create-a-load-balancer/).
- Update permission on the Load Balancer service to attach and detach, and create permission to make
  a new load balancer. See [TIR Access Control](/tir/tir-iam/).
- The port your application listens on. The VMs must answer on it.

## Make a load balancer for the group

On the **Associated Load Balancers** tab of the group, select **Create Load Balancer**. The new load
balancer starts with this group as its backend.

## Attach the group to a load balancer

1. On the details page of the group, open the **Associated Load Balancers** tab, which has every Web
   traffic load balancer the group can use.
2. Open the row menu of the load balancer and select **Attach**.
3. Under **Backend group**, check the **Name**, **Protocol** and **Port**. TIR suggests
   `<group name>-backend` as the name, adding -2, -3 and so on if the load balancer already uses it.
   Choose **HTTP**, **HTTPS**, or **gRPC**, and set **Port** to the port your application listens on.
4. Under **Routing rule**, check the **Rule name**, `<group name>-route` by default, and **When a
   request matches**. A rule with no condition sends every request to the group.
5. (Optional) Under **Advanced settings**, change how requests are distributed, the health check,
   the request timeout, and the number of retries.
6. Select **Attach**.

If the load balancer already has a rule with no condition, this group would receive nothing, so add
a condition to the new rule before you attach.

Attaching creates a backend group and a routing rule. If the load balancer refuses the rule, the
group is added but no traffic reaches it yet; detach, and attach again once the load balancer is
healthy.

### Attach from the load balancer

On the load balancer's own create or edit page, the backend picker lists an Auto Scaling Group under the
tab **Auto Scaling Group**. You can pick a group only while it is **Running**. This attaches the group itself,
not a snapshot of its VMs. An Auto Scaling Group needs a backend
group of its own that holds only the group.

## Check which load balancers serve the group

On **Associated Load Balancers**, **Traffic to this group** says whether requests reach the group.
**Serving** means they do. Otherwise it says why not: the route is missing and the load balancer
answers 503, the group has no VMs to serve, or the load balancer has failed, is stopped, is starting
or is unavailable. **Domain** is the load balancer's address, once it has one. An attached load
balancer stays listed even while it is **Initializing** or **Failed**. If there is none to attach,
create one from the same tab.

Once traffic reaches the group, **Connect** on **Overview** gives the load balancer's address as
`https://<domain>`.

## Detach the group

1. On the **Associated Load Balancers** tab, open the row menu of the attached load balancer and select
   **Detach**.
2. Confirm. The group's VMs stop receiving traffic from the load balancer and keep running, and you
   can attach the group again at any time.

Detaching also removes the backend group and its route when this group was its only member. If that
was the load balancer's only routing rule, it then answers 503 to every request.

## Check that traffic reaches the group

**Traffic to this group** reads **Serving** for the load balancer, and the **Domain** opens your
application. Requests are shared among the VMs that the group has running.

To see which VMs receive traffic, see [Monitor an Auto Scaling
Group](/tir/tir-auto-scaling/how-to/monitor-an-auto-scaling-group/). For help with a group that returns
503, see [Troubleshooting TIR Auto Scaling Groups](/tir/tir-auto-scaling/troubleshooting/tir-auto-scaling/).