Deploying for high availability

This page describes how to run the enterprise-store charm with multiple units for high availability, fronted by the haproxy charm which terminates client TLS at the edge and load-balances across the units.

It applies to both deployment modes: first deploy the charm as described in the Proxy mode or Offline mode guide, then add the haproxy integration and scale out as shown here. The snippets below show only the haproxy-specific parts, to be combined with the base deployment.

TLS options

If the registered store uses https, TLS can be handled in two ways:

• Let the charm terminate TLS, by setting the enterprise-store certificate / private_key config options (or set both to selfsigned for a self-signed certificate). This is suitable for a single-unit deployment.
• Put the haproxy charm in front of it via the haproxy-route interface, in which case haproxy terminates TLS at the edge and the enterprise-store is served over plain HTTP behind it. This is the recommended option for a multi-unit, highly available deployment.

Fronting with HAProxy

The charm can be placed behind the haproxy charm (channel 2.8/stable) using the haproxy-route interface. In this mode haproxy terminates client TLS at the edge and the enterprise-store snap is served over plain HTTP behind it. The snap honours the X-Forwarded-Proto header, so it still advertises https:// URLs to clients.

Two constraints apply when fronting with haproxy:

• The registration bundle’s domain must be haproxy’s public hostname, and haproxy’s edge certificate must cover that hostname. Clients connect to whatever domain is in the bundle.
• Do not set the enterprise-store certificate / private_key config options while related via haproxy-route. The haproxy charm owns TLS termination in this mode so setting them is contradictory. In the case these charm configs are set, the integration of the haproxy-route will unset them at the snap level and the stale certificate / private_key configs will need to be unset manually.

Scaling out

The charm supports multiple units for high availability; deploy with num_units / units greater than one, or scale later:

juju add-unit enterprise-store

Configuration, certificate and snap changes are coordinated across units through the rollingops-peers peer relation so that at most one unit restarts its services at a time. See the upgrading page for more detail on rolling operations.

Deploy with the Juju CLI

In addition to the base deployment, deploy haproxy with a certificate provider, relate both, and scale the enterprise-store to multiple units:

juju deploy haproxy --channel 2.8/stable
juju deploy self-signed-certificates
juju integrate haproxy:certificates self-signed-certificates
juju integrate enterprise-store:haproxy-route haproxy:haproxy-route
juju add-unit enterprise-store -n 2

Deploy with a Juju bundle

Add the haproxy-specific applications and relations to your deployment bundle, and set the enterprise-store to multiple units:

applications:
  haproxy:
    charm: haproxy
    channel: 2.8/stable
    num_units: 1
  self-signed-certificates:
    charm: self-signed-certificates
    num_units: 1
  enterprise-store:
    num_units: 3

relations:
- - haproxy:certificates
  - self-signed-certificates:certificates
- - enterprise-store:haproxy-route
  - haproxy:haproxy-route

Deploy with Terraform

Add the haproxy-specific resources to the module containing your base deployment. They reference the enterprise-store application and the model defined there (replace <model> with your model resource name):

resource "juju_application" "self-signed-certificates" {
  name       = "self-signed-certificates"
  model_uuid = juju_model.<model>.uuid
  charm {
    name = "self-signed-certificates"
  }
}

resource "juju_application" "haproxy" {
  name       = "haproxy"
  model_uuid = juju_model.<model>.uuid
  charm {
    name    = "haproxy"
    channel = "2.8/stable"
  }
}

resource "juju_integration" "haproxy-self-signed" {
  model_uuid = juju_model.<model>.uuid
  application {
    name     = juju_application.haproxy.name
    endpoint = "certificates"
  }
  application {
    name     = juju_application.self-signed-certificates.name
    endpoint = "certificates"
  }
}

resource "juju_integration" "haproxy-route" {
  model_uuid = juju_model.<model>.uuid
  application {
    name     = juju_application.haproxy.name
    endpoint = "haproxy-route"
  }
  application {
    name     = juju_application.enterprise-store.name
    endpoint = "haproxy-route"
  }
}

Set units = 3 (or higher) on the enterprise-store application in your base deployment to run multiple units behind haproxy.