data/aws/vpc: Use HTTPS for load balancer health checks

As suggested by Dani Comnea [1].  When we switched to network load
balancers in 16dfbb3 (data/aws: use nlbs instead of elbs,
2018-11-01, openshift#594), we replaced things like:

  resource "aws_elb" "api_internal" {
    ...
    health_check {
      healthy_threshold   = 2
      unhealthy_threshold = 2
      timeout             = 3
      target              = "SSL:6443"
      interval            = 5
    }
    ...
  }

with:

  resource "aws_lb_target_group" "api_internal" {
    ...
    health_check {
      healthy_threshold   = 3
      unhealthy_threshold = 3
      interval            = 10
      port                = 6443
      protocol            = "TCP"
    }
  }

This resulted in logs like [2]:

  [core@ip-10-0-11-88 ~]$ sudo crictl ps
  CONTAINER ID        IMAGE                                                                                                                                           CREATED             STATE               NAME                    ATTEMPT
  1bf4870ea6eea       registry.svc.ci.openshift.org/openshift/origin-v4.0-2018-12-15-160933@sha256:97eac256dde260e8bee9a5948efce5edb879dc6cb522a0352567010285378a56   2 minutes ago       Running             machine-config-server   0
  [core@ip-10-0-11-88 ~]$ sudo crictl logs 1bf4870ea6eea
  I1215 20:23:07.088210       1 bootstrap.go:37] Version: 3.11.0-356-gb7ffe0c7-dirty
  I1215 20:23:07.088554       1 api.go:54] launching server
  I1215 20:23:07.088571       1 api.go:54] launching server
  2018/12/15 20:24:17 http: TLS handshake error from 10.0.20.86:28372: EOF
  2018/12/15 20:24:18 http: TLS handshake error from 10.0.20.86:38438: EOF
  2018/12/15 20:24:18 http: TLS handshake error from 10.0.47.69:26320: EOF
  ...

when the health check opens a TCP connection (in this case to the
machine-config server on 49500) and then hangs up without completing
the TLS handshake.  Network load balancers [3,4] do not have an analog
to the classic load balancers' SSL protocol [5,6,7], so we're using
HTTPS.

There's some discussion in [8] about the best way to perform
unauthenticated liveness checks on the Kubernetes API server.  For
now, I'm assuming that both 200 and 401 responses to /healthz requests
indicate a functional server, and we can evaluate other response
status codes as necessary.  Checking against a recent cluster:

  $ curl -i https://wking-api.devcluster.openshift.com:6443/healthz
  curl: (60) Peer's Certificate issuer is not recognized.
  More details here: http://curl.haxx.se/docs/sslcerts.html

  curl performs SSL certificate verification by default, using a "bundle"
   of Certificate Authority (CA) public keys (CA certs). If the default
   bundle file isn't adequate, you can specify an alternate file
   using the --cacert option.
  If this HTTPS server uses a certificate signed by a CA represented in
   the bundle, the certificate verification probably failed due to a
   problem with the certificate (it might be expired, or the name might
   not match the domain name in the URL).
  If you'd like to turn off curl's verification of the certificate, use
   the -k (or --insecure) option.
  $ curl -ik https://wking-api.devcluster.openshift.com:6443/healthz
  HTTP/1.1 200 OK
  Cache-Control: no-store
  Date: Sun, 16 Dec 2018 06:18:23 GMT
  Content-Length: 2
  Content-Type: text/plain; charset=utf-8

I don't know if the network load balancer health checks care about
certificate validity or not.  I guess we'll see how CI testing handles
this.

Ignition is only exposed inside the cluster, and checking that from a
master node:

  [core@ip-10-0-26-134 ~]$ curl -i https://wking-api.devcluster.openshift.com:49500/
  curl: (60) Peer's Certificate issuer is not recognized.
  More details here: http://curl.haxx.se/docs/sslcerts.html

  curl performs SSL certificate verification by default, using a "bundle"
   of Certificate Authority (CA) public keys (CA certs). If the default
   bundle file isn't adequate, you can specify an alternate file
   using the --cacert option.
  If this HTTPS server uses a certificate signed by a CA represented in
   the bundle, the certificate verification probably failed due to a
   problem with the certificate (it might be expired, or the name might
   not match the domain name in the URL).
  If you'd like to turn off curl's verification of the certificate, use
   the -k (or --insecure) option.
  [core@ip-10-0-26-134 ~]$ curl -ik https://wking-api.devcluster.openshift.com:49500/
  HTTP/1.1 404 Not Found
  Content-Type: text/plain; charset=utf-8
  X-Content-Type-Options: nosniff
  Date: Sun, 16 Dec 2018 06:30:14 GMT
  Content-Length: 19

  404 page not found

So I've allowed 200, 401, and 404 there just to be generous.

[1]: openshift#923
[2]: https://groups.google.com/d/msg/openshift-4-dev-preview/Jmt6AK0EJR4/Ed3W7yZyBQAJ
[3]: https://docs.aws.amazon.com/elasticloadbalancing/latest/network/target-group-health-checks.html
[4]: https://www.terraform.io/docs/providers/aws/r/lb_target_group.html#protocol
[5]: https://docs.aws.amazon.com/elasticloadbalancing/latest/classic/elb-healthchecks.html
[6]: https://www.terraform.io/docs/providers/aws/r/elb.html#target
[7]: hashicorp/terraform-provider-aws#6866
[8]: kubernetes/kubernetes#43784

data/data/aws/vpc/master-elb.tf

@@ -44,7 +44,9 @@ resource "aws_lb_target_group" "api_internal" {
     unhealthy_threshold = 3
     interval            = 10
     port                = 6443
-    protocol            = "TCP"
+    protocol            = "HTTPS"
+    path                = "/healthz"
+    matcher             = "200,401"
   }
 }
 
@@ -66,7 +68,9 @@ resource "aws_lb_target_group" "api_external" {
     unhealthy_threshold = 3
     interval            = 10
     port                = 6443
-    protocol            = "TCP"
+    protocol            = "HTTPS"
+    path                = "/healthz"
+    matcher             = "200,401"
   }
 }
 
@@ -88,7 +92,9 @@ resource "aws_lb_target_group" "services" {
     unhealthy_threshold = 3
     interval            = 10
     port                = 49500
-    protocol            = "TCP"
+    protocol            = "HTTPS"
+    path                = "/"
+    matcher             = "200,401,404"
   }
 }