rshriram/nginx-upsync-demo

Simple demo with nginx-upsync-module and nginx-stream-upsync-module

nginx-upsync-demo

This repository contains a compiled version of nginx 1.11.5 with the http upsync module and the stream upsync module that is capable of updating the upstreams dynamically from Consul.

Requirements

• Docker Compose
• Python 2.7

Usage

./run_services.sh

This should start a simple setup with a nginx gateway and two backend microservices helloworld (a HTTP service) and hellotcp (a TCP service) each with two upstream instances. Each instance represents a different version of the same backend service.

You can now access http://localhost:8080/helloworld/hello for the helloworld HTTP microservice. The output would alterntate between two versions of the helloworld (v1 and v2).

$ curl http://localhost:8080/helloworld/hello
Hello version: v2, container: 5cb1b788a89a
$ curl http://localhost:8080/helloworld/hello
Hello version: v1, container: 39b6ba6bc1c6
$ curl http://localhost:8080/helloworld/hello
Hello version: v2, container: 5cb1b788a89a
$ curl http://localhost:8080/helloworld/hello
Hello version: v1, container: 39b6ba6bc1c6
$ curl http://localhost:8080/helloworld/hello

To access the hellotcp TCP based microservice, run ./tcp-echo-client.py localhost 9080. The output would alternate between two versions of the hellotcp service (v1 and v2).

$ ./tcp-echo-client.py localhost 9080
hellotcp, version v2, host 89f3621d9b7a. Echoing hello there
$ ./tcp-echo-client.py localhost 9080
hellotcp, version v1, host 80fc4a874ac5. Echoing hello there
$ ./tcp-echo-client.py localhost 9080
hellotcp, version v2, host 89f3621d9b7a. Echoing hello there
$ ./tcp-echo-client.py localhost 9080
hellotcp, version v1, host 80fc4a874ac5. Echoing hello there

To see list of upstreams in nginx,

# For HTTP services
$ curl http://localhost:8080/upstream_show
Upstream name: helloworld; Backend server count: 2
        server 172.17.0.6:5000 weight=1 max_fails=2 fail_timeout=10s;
        server 172.17.0.3:5000 weight=1 max_fails=2 fail_timeout=10s;

# For TCP services
$ curl http://localhost:9180/upstream_show
Upstream name: hellotcp; Backend server count: 2
        server 172.17.0.5:5000 weight=1 max_fails=2 fail_timeout=10s;
        server 172.17.0.4:5000 weight=1 max_fails=2 fail_timeout=10s;

Changing the weights

In order to increase the weights for a particular version of the service (e.g., helloworld v2),

• Get the IP address of helloworld-v2 service using the following command

helloworldv2=`docker inspect helloworld-v2|grep -w IPAddress|tail -1|cut -d : -f2|tr -d ' ",'`

• Update Consul's entry for helloworld-v2 with appropriate weights (this gets translated into nginx's weights for upstream instances automatically).

curl -X PUT -d "{\"weight\":3, \"max_fails\":2, \"fail_timeout\":10}" http://localhost:8500/v1/kv/helloworld/${helloworldv2}:5000

• Access helloworld

curl http://localhost:8080/helloworld/hello

You should see the traffic being split into 25% for helloworld v1 and 75% for helloworld v2. For more details on adjusting weights, deleting upstream instances, etc., refer to the nginx-upsync module documentation