deploy.rst

Deployment

EMQ X Cluster can be deployed as an IoT Hub on an enterprise's private cloud, or on public clouds such as AWS, Azure and QingCloud.

Typical deployment architecture:

Load Balancer (LB)

The Load Balancer (LB) distributes MQTT connections and traffic from devices across the EMQ X clusters. LB enhances the HA of the clusters, balances the loads among the cluster nodes and makes the dynamic expansion possible.

It is recommended that SSL connections are terminated by a LB. The links between devices and the LB are secured by SSL, while the links between the LB and EMQ X cluster nodes are plain TCP connections. By this setup, a single EMQ X cluster can serve a million devices.

LB products of public cloud providers:

Cloud provider	SSL Termination	LB Product DOC/URL
QingCloud	Y	https://docs.qingcloud.com/guide/loadbalancer.html
AWS	Y	https://aws.amazon.com/cn/elasticloadbalancing/
aliyun	N	https://www.aliyun.com/product/slb

LBs for Private Cloud:

Open-Source LB	SSL Termination	DOC/URL
HAProxy	Y	https://www.haproxy.com/solutions/load-balancing.html
NGINX	PLUS Edition	https://www.nginx.com/solutions/load-balancing/

Recommend AWS with ELB for a public cloud deployment, and HAProxy for a private cloud deployment.

EMQ X Cluster

EMQ X cluster nodes are deployed behind LB. It is suggested that the nodes are deployed on VPCs or on a private network. Cloud provider, such as AWS, Azure or QingCloud, usually provides VPC network.

EMQ X Provides the MQTT service on following TCP ports by default:

1883	MQTT
8883	MQTT/SSL
8083	MQTT/WebSocket
8084	MQTT/WebSocket/SSL
8080	Management API
18083	Dashboard

Firewall should make the relevant ports accessible for public.

Following ports are opened for cluster internal communication:

4369	Node discovery port
5369	Cluster PRC
6369	Cluster channel

If deployed between nodes, firewalls should be configured that the above ports are inter-accessible between the nodes.

Deploying on QingCloud

1. Create VPC network.
2. Create a 'private network' for EMQ X cluster inside the VPC network, e.g. 192.168.0.0/24
3. Create 2 EMQ X hosts inside the private network, like:

emqx1	192.168.0.2
emqx2	192.168.0.3

4. Install and cluster EMQ X on these two hosts. Please refer to the sections of cluster installation for details.
5. Create LB and assign the public IP address.
6. Create MQTT TCP listener:

Or create SSL listener and terminate the SSL connections on LB:

7. Connect the MQTT clients to the LB using the public IP address and test the deployment.

Deploying on AWS

1. Create VPC network.
2. Create a 'private network' for EMQ X cluster inside the VPC network, e.g. 192.168.0.0/24
3. Create 2 hosts inside the private network, like:

emqx1	192.168.0.2
emqx2	192.168.0.3

4. Open the TCP ports for MQTT services (e.g. 1883,8883) on the security group.
5. Install and cluster EMQ X on these two hosts. Please refer to the sections of cluster installation for details.
6. Create ELB (Classic Load Balancer), assign the VPC network, and assign the public IP address.
7. Create MQTT TCP listener on the ELB:

Or create SSL listener and terminate the SSL connections on the ELB:

8. Connect the MQTT clients to the ELB using the public IP address and test the deployment.

Deploying on private network

Direct connection of EMQ X cluster

EMQ X cluster should be DNS-resolvable and the clients access the cluster via domain name or IP list:

1. Deploy EMQ X cluster. Please refer to the sections of 'Installation' and 'EMQ X nodes clustering' for details.
2. Enable the access to the MQTT ports on the firewall (e.g. 1883, 8883).
3. Client devices access the EMQ X cluster via domain name or IP list.

Note

This kind of deployment is NOT recommended.

HAProxy -> EMQ X Cluster

HAProxy serves as a LB for EMQ X cluster and terminates the SSL connections:

1. Create EMQ X Cluster nodes like following:

node	IP
emqx1	192.168.0.2
emqx2	192.168.0.3

2. Modify the /etc/haproxy/haproxy.cfg accordingly. An example:

   listen mqtt-ssl
       bind *:8883 ssl crt /etc/ssl/emqx/emqx.pem no-sslv3
       mode tcp
       maxconn 50000
       timeout client 600s
       default_backend emqx_cluster
   
   backend emqx_cluster
       mode tcp
       balance source
       timeout server 50s
       timeout check 5000
       server emqx1 192.168.0.2:1883 check inter 10000 fall 2 rise 5 weight 1
       server emqx2 192.168.0.3:1883 check inter 10000 fall 2 rise 5 weight 1
       source 0.0.0.0 usesrc clientip
   

NGINX Plus -> EMQ X Cluster

NGINX Plus serves as a LB for EMQ X cluster and terminates the SSL connections:

1. Install the NGINX Plus. An instruction for Ubuntu: https://cs.nginx.com/repo_setup
2. Create EMQ X cluster nodes like following:

node	IP
emqx1	192.168.0.2
emqx2	192.168.0.3

3. Modify the /etc/nginx/nginx.conf. An example:

   stream {
       # Example configuration for TCP load balancing
   
       upstream stream_backend {
           zone tcp_servers 64k;
           hash $remote_addr;
           server 192.168.0.2:1883 max_fails=2 fail_timeout=30s;
           server 192.168.0.3:1883 max_fails=2 fail_timeout=30s;
       }
   
       server {
           listen 8883 ssl;
           status_zone tcp_server;
           proxy_pass stream_backend;
           proxy_buffer_size 4k;
           ssl_handshake_timeout 15s;
           ssl_certificate     /etc/emqx/certs/cert.pem;
           ssl_certificate_key /etc/emqx/certs/key.pem;
       }
   }