Ok, problem solved. When I used LoadBalancer as a service I didn’t have access to its IP from within terraform. However, when using ingress-nginx helm chart I have the IP. Yay!

So the solution is:

# 1. Download ingress-nginx chart
resource "helm_release" "ingress-nginx" {
  repository = "https://kubernetes.github.io/ingress-nginx"
  chart = "ingress-nginx"
  name = "ingress-nginx"
}

# 2. Add a Service with `type = "ClusterIP"`, which would expose the app to the Kubernetes cluster
resource "kubernetes_service" "blog-backend" {
  metadata {
    name = "blog-backend-service"
    labels = {
      app = "blog-backend"
    }
  }
  spec {
    type = "ClusterIP"

    selector = {
      app = "blog-backend"
    }

    port {
      port = 8080
      target_port = 8080
    }
  }
}

# 3. Create an ingress controller, which maps a domain to a service above
resource "kubernetes_ingress" "blog-backend-ingress" {
  metadata {
    name = "blog-backend-ingress"
    annotations = {
      "kubernetes.io/ingress.class" = "nginx"
    }
  }
  spec {
    rule {
      host = "blog.mydomain.com"
      http {
        path {
          backend {
            service_name = "blog-backend-service"
            service_port = 8080
          }
          path = "/"
        }
      }
    }
  }
}

# 4. Create a domain
resource "digitalocean_domain" "blog-backend-domain" {
  name = "blog.mydomain.com"
}

# 5. Create a domain record, whose IP (`value`) we get from the ingress from step 3.
resource "digitalocean_record" "backend-api" {
  domain = digitalocean_domain.blog-backend-domain.name
  name = "@"
  type = "A"
  value = kubernetes_ingress.blog-backend-ingress.load_balancer_ingress[0].ip
}