Connect an Auto Scaling Group

Connect an Auto Scaling Group to a Load Balancer so that the Load Balancer sends traffic to the group’s VMs. The Load Balancer is attached to the group itself, so traffic keeps following the group through every scale-up and scale-down. Only a web Load Balancer can have an Auto Scaling Group as a backend.

Before you begin

• A web Load Balancer. To make one, see Create a Load Balancer.
• An Auto Scaling Group in the Running state in the same project.
• Permission to update Load Balancers in the project. See TIR Access Control.

Connect from the Load Balancer

1. Open the Load Balancer and select the Backend groups tab.
2. Select New group. To change a group that exists, open Edit from its Actions menu.
3. Enter a Name, then choose the Auto Scaling Group tab among the backends. It lists the project’s running Auto Scaling Groups, each with its name, how many VMs it is running, and the range it scales between.
4. Tick the group. The Auto Scaling Group is one backend, with a fixed port of 80 and a fixed weight of 1.
5. Select Create group.

A backend group holds one kind of backend, so VMs and inference endpoints go in a separate group from an Auto Scaling Group. For the other settings of the group, see Create a backend group. Traffic reaches the group once a rule forwards to it; see Add a rule.

Connect from the Auto Scaling Group

On the Auto Scaling Group’s details page, the Associated Load Balancers tab has the project’s web Load Balancers. A Load Balancer that is Initializing or Failed is listed only when the group is already attached to it.

1. In the row of the Load Balancer, open the menu and select Attach.
2. Under Backend group, check the name, Protocol and Port. The backend group is named after the group with “-backend” added. Choose HTTP, the default, HTTPS or gRPC. Port starts at 80; set it to the port your application listens on.
3. Check the Routing rule section. The rule is named after the group with “-route” added. It matches every request, so it sends 100% of requests to the new backend group.
4. (Optional) Open Advanced settings to set Distribute requests by, the health check, Request timeout (s) and Retry attempts.
5. Select Attach. The group is now behind the Load Balancer.

Both names are required, the backend group name must be new on that Load Balancer, and the port must be valid. If the Load Balancer already has a rule that matches every request, the first matching rule wins, so add a condition to the new rule.

If the backend group is created but the rule is refused, traffic has yet to reach the group. Detach the group and try again once the Load Balancer is healthy.

To start from the group instead, select Create Load Balancer on the same tab. The new Load Balancer starts with the group as a backend.

Check that traffic reaches the group

Traffic to this group says whether the Load Balancer is serving the group. Not routed means the Load Balancer is Running and still needs an active rule that forwards to the group’s backend group. Add a rule that forwards to it.

Detach the group

Detach an Auto Scaling Group to stop the Load Balancer sending traffic to it. In the row of the Load Balancer, open the menu, select Detach, and confirm. The group’s VMs stop receiving traffic from the Load Balancer immediately. They keep running, and you can attach them again.

Detaching removes the group from every backend group of that Load Balancer:

• A backend group that held only this Auto Scaling Group is deleted, and so is a rule that forwarded only to it.
• A rule that also forwards to other backend groups only loses this destination.

Before you confirm, the dialog names the backend groups that will be removed along with their routes. When the detach would remove the Load Balancer’s only rule, the dialog also warns that the Load Balancer will then answer 503 to every request.

To make the same connection from the group’s side, see Connect a load balancer. To control which traffic can reach the Load Balancer next, see Manage security groups.