# AWS Load Balancers

To distribute traffic to the Octopus web portal on multiple nodes, you need to use a HTTP load balancer. AWS provides a solution to distribute HTTP/HTTPS traffic to EC2 instances, Elastic Load Balancing is a highly available, secure, and elastic load balancer. There are three implementations of ELB;

- [Application Load Balancer](https://docs.aws.amazon.com/elasticloadbalancing/latest/application/introduction.html)
- [Network Load Balancer](https://docs.aws.amazon.com/elasticloadbalancing/latest/network/introduction.html)
- [Classic Load Balancer](https://docs.aws.amazon.com/elasticloadbalancing/latest/classic/introduction.html)
- [Comparison Table](https://aws.amazon.com/elasticloadbalancing/features/#Product_comparisons)

## Tentacle

If you are *only* using [Listening Tentacles](/docs/infrastructure/deployment-targets/tentacle/tentacle-communication/#listening-tentacles-recommended), we recommend using the Application Load Balancer.

However, [Polling Tentacles](/docs/infrastructure/deployment-targets/tentacle/tentacle-communication/#polling-tentacles) don't work well with the Application Load Balancer, so instead, we recommend using the Network Load Balancer. To setup a Network Load Balancer for Octopus High Availability with Polling Tentacles take a look at this [knowledge base article](https://help.octopus.com/t/how-can-i-configure-my-polling-tentacles-to-hit-my-octopus-deploy-high-availability-instance-to-sitting-behind-an-aws-load-balancer/24890).

## gRPC

gRPC traffic can be routed with either the Application Load Balancer which provides first class support for gRPC or with the Network/Classic Load Balancer.

If you choose to use the Application Load Balancer you might come across certain errors like the following from our gRPC clients(Kubernetes Monitor/Argo CD Gateway)

```go
stream terminated by RST_STREAM with error code: PROTOCOL_ERROR
```

This is due to how AWS's Application Load Balancer ignores HTTP/2 PING Frames which causes the long lived gRPC streams we use to be ended after 60 seconds of idle time by default. [See the AWS docs for more details](https://docs.aws.amazon.com/elasticloadbalancing/latest/application/edit-load-balancer-attributes.html#connection-idle-timeout).

To prevent this error from occurring the Application Load Balancer needs to be configured with an increased idle timeout. This can be done by setting the following attribute `idle_timeout.timeout_seconds` to a value between 1-4000 seconds.