Permalink
497 lines (384 sloc) 25.6 KB

Nginx Ingress Controller

This is an nginx Ingress controller that uses ConfigMap to store the nginx configuration. See Ingress controller documentation for details on how it works.

Contents

Conventions

Anytime we reference a tls secret, we mean (x509, pem encoded, RSA 2048, etc). You can generate such a certificate with: openssl req -x509 -nodes -days 365 -newkey rsa:2048 -keyout ${KEY_FILE} -out ${CERT_FILE} -subj "/CN=${HOST}/O=${HOST}" and create the secret via kubectl create secret tls ${CERT_NAME} --key ${KEY_FILE} --cert ${CERT_FILE}

Requirements

Command line arguments

Usage of :
      --alsologtostderr                  log to standard error as well as files
      --apiserver-host string            The address of the Kubernetes Apiserver to connect to in the format of protocol://address:port, e.g., http://localhost:8080. If not specified, the assumption is that the binary runs inside a Kubernetes cluster and local discovery is attempted.
      --configmap string                 Name of the ConfigMap that contains the custom configuration use
      --default-backend-service string   Service used to serve a 404 page for the default backend. Takes the form namespace/name. The controller uses the first node port of this Service for the default backend.
      --default-ssl-certificate string   Name of the secret that contains a SSL certificate to be used as default for a HTTPS catch-all server
      --election-id string               Election id to use for status update. (default "ingress-controller-leader")
      --force-namespace-isolation        Force namespace isolation. This flag is required to avoid the reference of secrets or configmaps located in a different namespace than the specified in the flag --watch-namespace.
      --health-check-path string         Defines the URL to be used as health check inside in the default server in NGINX. (default "/healthz")
      --healthz-port int                 port for healthz endpoint. (default 10254)
      --ingress-class string             Name of the ingress class to route through this controller.
      --kubeconfig string                Path to kubeconfig file with authorization and master location information.
      --log_backtrace_at traceLocation   when logging hits line file:N, emit a stack trace (default :0)
      --log_dir string                   If non-empty, write log files in this directory
      --logtostderr                      log to standard error instead of files
      --profiling                        Enable profiling via web interface host:port/debug/pprof/ (default true)
      --publish-service string           Service fronting the ingress controllers. Takes the form namespace/name. The controller will set the endpoint records on the ingress objects to reflect those on the service.
      --stderrthreshold severity         logs at or above this threshold go to stderr (default 2)
      --sync-period duration             Relist and confirm cloud resources this often. (default 1m0s)
      --tcp-services-configmap string    Name of the ConfigMap that contains the definition of the TCP services to expose.
		  The key in the map indicates the external port to be used. The value is the name of the service with the format namespace/serviceName and the port of the service could be a number of the name of the port.
		  The ports 80 and 443 are not allowed as external ports. This ports are reserved for the backend
      --udp-services-configmap string    Name of the ConfigMap that contains the definition of the UDP services to expose.
		  The key in the map indicates the external port to be used. The value is the name of the service with the format namespace/serviceName and the port of the service could be a number of the name of the port.
      --update-status                    Indicates if the ingress controller should update the Ingress status IP/hostname. Default is true (default true)
-v, --v Level                          log level for V logs
      --vmodule moduleSpec               comma-separated list of pattern=N settings for file-filtered logging
      --watch-namespace string           Namespace to watch for Ingress. Default is to watch all namespaces

Try running the Ingress controller

Before deploying the controller to production you might want to run it outside the cluster and observe it.

$ make build
$ mkdir /etc/nginx-ssl
$ ./rootfs/nginx-ingress-controller --running-in-cluster=false --default-backend-service=kube-system/default-http-backend

Deployment

First create a default backend:

$ kubectl create -f examples/deployment/nginx/default-backend.yaml
$ kubectl expose rc default-http-backend --port=80 --target-port=8080 --name=default-http-backend

Loadbalancers are created via a ReplicationController or Daemonset:

$ kubectl create -f examples/default/rc-default.yaml

HTTP

First we need to deploy some application to publish. To keep this simple we will use the echoheaders app that just returns information about the http request as output

kubectl run echoheaders --image=gcr.io/google_containers/echoserver:1.5 --replicas=1 --port=8080

Now we expose the same application in two different services (so we can create different Ingress rules)

kubectl expose deployment echoheaders --port=80 --target-port=8080 --name=echoheaders-x
kubectl expose deployment echoheaders --port=80 --target-port=8080 --name=echoheaders-y

Next we create a couple of Ingress rules

kubectl create -f examples/ingress.yaml

we check that ingress rules are defined:

$ kubectl get ing
NAME      RULE          BACKEND   ADDRESS
echomap   -
          foo.bar.com
          /foo          echoheaders-x:80
          bar.baz.com
          /bar          echoheaders-y:80
          /foo          echoheaders-x:80

Before the deploy of the Ingress controller we need a default backend 404-server

kubectl create -f examples/default-backend.yaml
kubectl expose rc default-http-backend --port=80 --target-port=8080 --name=default-http-backend

Check NGINX it is running with the defined Ingress rules:

$ LBIP=$(kubectl get node `kubectl get po -l name=nginx-ingress-lb --template '{{range .items}}{{.spec.nodeName}}{{end}}'` --template '{{range $i, $n := .status.addresses}}{{if eq $n.type "ExternalIP"}}{{$n.address}}{{end}}{{end}}')
$ curl $LBIP/foo -H 'Host: foo.bar.com'

HTTPS

You can secure an Ingress by specifying a secret that contains a TLS private key and certificate. Currently the Ingress only supports a single TLS port, 443, and assumes TLS termination. This controller supports SNI. The TLS secret must contain keys named tls.crt and tls.key that contain the certificate and private key to use for TLS, eg:

apiVersion: v1
data:
  tls.crt: base64 encoded cert
  tls.key: base64 encoded key
kind: Secret
metadata:
  name: foo-secret
  namespace: default
type: kubernetes.io/tls

Referencing this secret in an Ingress will tell the Ingress controller to secure the channel from the client to the loadbalancer using TLS:

apiVersion: extensions/v1beta1
kind: Ingress
metadata:
  name: no-rules-map
spec:
  tls:
    secretName: foo-secret
  backend:
    serviceName: s1
    servicePort: 80

Please follow test.sh as a guide on how to generate secrets containing SSL certificates. The name of the secret can be different than the name of the certificate.

Check the example

Default SSL Certificate

NGINX provides the option server name as a catch-all in case of requests that do not match one of the configured server names. This configuration works without issues for HTTP traffic. In case of HTTPS NGINX requires a certificate. For this reason the Ingress controller provides the flag --default-ssl-certificate. The secret behind this flag contains the default certificate to be used in the mentioned case. If this flag is not provided NGINX will use a self signed certificate.

Running without the flag --default-ssl-certificate:

$ curl -v https://10.2.78.7:443 -k
* Rebuilt URL to: https://10.2.78.7:443/
*   Trying 10.2.78.4...
* Connected to 10.2.78.7 (10.2.78.7) port 443 (#0)
* ALPN, offering http/1.1
* Cipher selection: ALL:!EXPORT:!EXPORT40:!EXPORT56:!aNULL:!LOW:!RC4:@STRENGTH
* successfully set certificate verify locations:
*   CAfile: /etc/ssl/certs/ca-certificates.crt
  CApath: /etc/ssl/certs
* TLSv1.2 (OUT), TLS header, Certificate Status (22):
* TLSv1.2 (OUT), TLS handshake, Client hello (1):
* TLSv1.2 (IN), TLS handshake, Server hello (2):
* TLSv1.2 (IN), TLS handshake, Certificate (11):
* TLSv1.2 (IN), TLS handshake, Server key exchange (12):
* TLSv1.2 (IN), TLS handshake, Server finished (14):
* TLSv1.2 (OUT), TLS handshake, Client key exchange (16):
* TLSv1.2 (OUT), TLS change cipher, Client hello (1):
* TLSv1.2 (OUT), TLS handshake, Finished (20):
* TLSv1.2 (IN), TLS change cipher, Client hello (1):
* TLSv1.2 (IN), TLS handshake, Finished (20):
* SSL connection using TLSv1.2 / ECDHE-RSA-AES128-GCM-SHA256
* ALPN, server accepted to use http/1.1
* Server certificate:
*    subject: CN=foo.bar.com
*    start date: Apr 13 00:50:56 2016 GMT
*    expire date: Apr 13 00:50:56 2017 GMT
*    issuer: CN=foo.bar.com
*    SSL certificate verify result: self signed certificate (18), continuing anyway.
> GET / HTTP/1.1
> Host: 10.2.78.7
> User-Agent: curl/7.47.1
> Accept: */*
>
< HTTP/1.1 404 Not Found
< Server: nginx/1.11.1
< Date: Thu, 21 Jul 2016 15:38:46 GMT
< Content-Type: text/html
< Transfer-Encoding: chunked
< Connection: keep-alive
< Strict-Transport-Security: max-age=15724800; includeSubDomains; preload
<
<span>The page you're looking for could not be found.</span>

* Connection #0 to host 10.2.78.7 left intact

Specifying --default-ssl-certificate=default/foo-tls:

core@localhost ~ $ curl -v https://10.2.78.7:443 -k
* Rebuilt URL to: https://10.2.78.7:443/
*   Trying 10.2.78.7...
* Connected to 10.2.78.7 (10.2.78.7) port 443 (#0)
* ALPN, offering http/1.1
* Cipher selection: ALL:!EXPORT:!EXPORT40:!EXPORT56:!aNULL:!LOW:!RC4:@STRENGTH
* successfully set certificate verify locations:
*   CAfile: /etc/ssl/certs/ca-certificates.crt
  CApath: /etc/ssl/certs
* TLSv1.2 (OUT), TLS header, Certificate Status (22):
* TLSv1.2 (OUT), TLS handshake, Client hello (1):
* TLSv1.2 (IN), TLS handshake, Server hello (2):
* TLSv1.2 (IN), TLS handshake, Certificate (11):
* TLSv1.2 (IN), TLS handshake, Server key exchange (12):
* TLSv1.2 (IN), TLS handshake, Server finished (14):
* TLSv1.2 (OUT), TLS handshake, Client key exchange (16):
* TLSv1.2 (OUT), TLS change cipher, Client hello (1):
* TLSv1.2 (OUT), TLS handshake, Finished (20):
* TLSv1.2 (IN), TLS change cipher, Client hello (1):
* TLSv1.2 (IN), TLS handshake, Finished (20):
* SSL connection using TLSv1.2 / ECDHE-RSA-AES128-GCM-SHA256
* ALPN, server accepted to use http/1.1
* Server certificate:
*    subject: CN=foo.bar.com
*    start date: Apr 13 00:50:56 2016 GMT
*    expire date: Apr 13 00:50:56 2017 GMT
*    issuer: CN=foo.bar.com
*    SSL certificate verify result: self signed certificate (18), continuing anyway.
> GET / HTTP/1.1
> Host: 10.2.78.7
> User-Agent: curl/7.47.1
> Accept: */*
>
< HTTP/1.1 404 Not Found
< Server: nginx/1.11.1
< Date: Mon, 18 Jul 2016 21:02:59 GMT
< Content-Type: text/html
< Transfer-Encoding: chunked
< Connection: keep-alive
< Strict-Transport-Security: max-age=15724800; includeSubDomains; preload
<
<span>The page you're looking for could not be found.</span>

* Connection #0 to host 10.2.78.7 left intact

Server-side HTTPS enforcement

By default the controller redirects (301) to HTTPS if TLS is enabled for that ingress . If you want to disable that behaviour globally, you can use ssl-redirect: "false" in the NGINX config map.

To configure this feature for specific ingress resources, you can use the ingress.kubernetes.io/ssl-redirect: "false" annotation in the particular resource.

HTTP Strict Transport Security

HTTP Strict Transport Security (HSTS) is an opt-in security enhancement specified through the use of a special response header. Once a supported browser receives this header that browser will prevent any communications from being sent over HTTP to the specified domain and will instead send all communications over HTTPS.

By default the controller redirects (301) to HTTPS if there is a TLS Ingress rule.

To disable this behavior use hsts=false in the NGINX config map.

Automated Certificate Management with Kube-Lego

Kube-Lego automatically requests missing certificates or expired from Let's Encrypt by monitoring ingress resources and its referenced secrets. To enable this for an ingress resource you have to add an annotation:

kubectl annotate ing ingress-demo kubernetes.io/tls-acme="true"

To setup Kube-Lego you can take a look at this full example. The first version to fully support Kube-Lego is nginx Ingress controller 0.8.

Exposing TCP services

Ingress does not support TCP services (yet). For this reason this Ingress controller uses the flag --tcp-services-configmap to point to an existing config map where the key is the external port to use and the value is <namespace/service name>:<service port>:[PROXY] It is possible to use a number or the name of the port. The last field is optional. Adding PROXY in the last field we can enable Proxy Protocol in a TCP service.

The next example shows how to expose the service example-go running in the namespace default in the port 8080 using the port 9000

apiVersion: v1
kind: ConfigMap
metadata:
  name: tcp-configmap-example
data:
  9000: "default/example-go:8080"

Please check the tcp services example

Exposing UDP services

Since 1.9.13 NGINX provides UDP Load Balancing.

Ingress does not support UDP services (yet). For this reason this Ingress controller uses the flag --udp-services-configmap to point to an existing config map where the key is the external port to use and the value is <namespace/service name>:<service port> It is possible to use a number or the name of the port.

The next example shows how to expose the service kube-dns running in the namespace kube-system in the port 53 using the port 53

apiVersion: v1
kind: ConfigMap
metadata:
  name: udp-configmap-example
data:
  53: "kube-system/kube-dns:53"

Please check the udp services example

Proxy Protocol

If you are using a L4 proxy to forward the traffic to the NGINX pods and terminate HTTP/HTTPS there, you will lose the remote endpoint's IP addresses. To prevent this you could use the Proxy Protocol for forwarding traffic, this will send the connection details before forwarding the actual TCP connection itself.

Amongst others ELBs in AWS and HAProxy support Proxy Protocol.

Please check the proxy-protocol example

Custom errors

In case of an error in a request the body of the response is obtained from the default backend. Each request to the default backend includes two headers:

  • X-Code indicates the HTTP code
  • X-Format the value of the Accept header

Using this two headers is possible to use a custom backend service like this one that inspect each request and returns a custom error page with the format expected by the client. Please check the example custom-errors

NGINX status page

The ngx_http_stub_status_module module provides access to basic status information. This is the default module active in the url /nginx_status. This controller provides an alternative to this module using nginx-module-vts third party module. To use this module just provide a config map with the key enable-vts-status=true. The URL is exposed in the port 18080. Please check the example example/rc-default.yaml

nginx-module-vts screenshot

To extract the information in JSON format the module provides a custom URL: /nginx_status/format/json

Running multiple ingress controllers

If you're running multiple ingress controllers, or running on a cloudprovider that natively handles ingress, you need to specify the annotation kubernetes.io/ingress.class: "nginx" in all ingresses that you would like this controller to claim. Not specifying the annotation will lead to multiple ingress controllers claiming the same ingress. Specifying the wrong value will result in all ingress controllers ignoring the ingress. Multiple ingress controllers running in the same cluster was not supported in Kubernetes versions < 1.3.

Running on Cloudproviders

If you're running this ingress controller on a cloudprovider, you should assume the provider also has a native Ingress controller and specify the ingress.class annotation as indicated in this section. In addition to this, you will need to add a firewall rule for each port this controller is listening on, i.e :80 and :443.

Disabling NGINX ingress controller

Setting the annotation kubernetes.io/ingress.class to any value other than "nginx" or the empty string, will force the NGINX Ingress controller to ignore your Ingress. Do this if you wish to use one of the other Ingress controllers at the same time as the NGINX controller.

Log format

The default configuration uses a custom logging format to add additional information about upstreams

    log_format upstreaminfo '{{ if $cfg.useProxyProtocol }}$proxy_protocol_addr{{ else }}$remote_addr{{ end }} - '
        '[$proxy_add_x_forwarded_for] - $remote_user [$time_local] "$request" $status $body_bytes_sent "$http_referer" "$http_user_agent" '
        '$request_length $request_time [$proxy_upstream_name] $upstream_addr $upstream_response_length $upstream_response_time $upstream_status';

Sources:

Description:

  • $proxy_protocol_addr: if PROXY protocol is enabled
  • $remote_addr: if PROXY protocol is disabled (default)
  • $proxy_add_x_forwarded_for: the X-Forwarded-For client request header field with the $remote_addr variable appended to it, separated by a comma
  • $remote_user: user name supplied with the Basic authentication
  • $time_local: local time in the Common Log Format
  • $request: full original request line
  • $status: response status
  • $body_bytes_sent: number of bytes sent to a client, not counting the response header
  • $http_referer: value of the Referer header
  • $http_user_agent: value of User-Agent header
  • $request_length: request length (including request line, header, and request body)
  • $request_time: time elapsed since the first bytes were read from the client
  • $proxy_upstream_name: name of the upstream. The format is upstream-<namespace>-<service name>-<service port>
  • $upstream_addr: keeps the IP address and port, or the path to the UNIX-domain socket of the upstream server. If several servers were contacted during request processing, their addresses are separated by commas
  • $upstream_response_length: keeps the length of the response obtained from the upstream server
  • $upstream_response_time: keeps time spent on receiving the response from the upstream server; the time is kept in seconds with millisecond resolution
  • $upstream_status: keeps status code of the response obtained from the upstream server

Local cluster

Using hack/local-up-cluster.sh is possible to start a local kubernetes cluster consisting of a master and a single node. Please read running-locally.md for more details.

Use of hostNetwork: true in the ingress controller is required to falls back at localhost:8080 for the apiserver if every other client creation check fails (eg: service account not present, kubeconfig doesn't exist, no master env vars...)

Debug & Troubleshooting

Using the flag --v=XX it is possible to increase the level of logging. In particular:

  • --v=2 shows details using diff about the changes in the configuration in nginx
I0316 12:24:37.581267       1 utils.go:148] NGINX configuration diff a//etc/nginx/nginx.conf b//etc/nginx/nginx.conf
I0316 12:24:37.581356       1 utils.go:149] --- /tmp/922554809  2016-03-16 12:24:37.000000000 +0000
+++ /tmp/079811012  2016-03-16 12:24:37.000000000 +0000
@@ -235,7 +235,6 @@

     upstream default-echoheadersx {
         least_conn;
-        server 10.2.112.124:5000;
         server 10.2.208.50:5000;

     }
I0316 12:24:37.610073       1 command.go:69] change in configuration detected. Reloading...
  • --v=3 shows details about the service, Ingress rule, endpoint changes and it dumps the nginx configuration in JSON format
  • --v=5 configures NGINX in debug mode

These issues were encountered in past versions of Kubernetes:

1.2.0-alpha7 deployment:

  • make setup-files.sh file in hypercube does not provide 10.0.0.1 IP to make-ca-certs, resulting in CA certs that are issued to the external cluster IP address rather then 10.0.0.1 -> this results in nginx-third-party-lb appearing to get stuck at "Utils.go:177 - Waiting for default/default-http-backend" in the docker logs. Kubernetes will eventually kill the container before nginx-third-party-lb times out with a message indicating that the CA certificate issuer is invalid (wrong ip), to verify this add zeros to the end of initialDelaySeconds and timeoutSeconds and reload the RC, and docker will log this error before kubernetes kills the container.

Limitations

  • Ingress rules for TLS require the definition of the field host

Why endpoints and not services

The NGINX ingress controller does not uses Services to route traffic to the pods. Instead it uses the Endpoints API in order to bypass kube-proxy to allow NGINX features like session affinity and custom load balancing algorithms. It also removes some overhead, such as conntrack entries for iptables DNAT.

NGINX notes

Since gcr.io/google_containers/nginx-slim:0.8 NGINX contains the next patches:

  • Dynamic TLS record size nginx__dynamic_tls_records.patch NGINX provides the parameter ssl_buffer_size to adjust the size of the buffer. Default value in NGINX is 16KB. The ingress controller changes the default to 4KB. This improves the TLS Time To First Byte (TTTFB) but the size is fixed. This patches adapts the size of the buffer to the content is being served helping to improve the perceived latency.

  • Add SPDY support back to Nginx with HTTP/2 nginx_1_9_15_http2_spdy.patch At the same NGINX introduced HTTP/2 support for SPDY was removed. This patch add support for SPDY without compromising HTTP/2 support using the Application-Layer Protocol Negotiation (ALPN) or Next Protocol Negotiation (NPN) Transport Layer Security (TLS) extension to negotiate what protocol the server and client support

openssl s_client -servername www.my-site.com -connect www.my-site.com:443 -nextprotoneg ''
CONNECTED(00000003)
Protocols advertised by server: h2, spdy/3.1, http/1.1