| title | Deployment with Docker | ||||
|---|---|---|---|---|---|
| slug | deployment-with-docker | ||||
| excerpt | |||||
| hidden | false | ||||
| metadata |
|
||||
| createdAt | Mon Jul 13 2015 09:27:33 GMT+0000 (Coordinated Universal Time) | ||||
| updatedAt | Thu Feb 08 2018 09:53:19 GMT+0000 (Coordinated Universal Time) |
The easiest way to try DeviceHive locally or in your development datacenter is to deploy it using Docker Compose.
This will start following basic DeviceHive services:
- DeviceHive Frontend service
- DeviceHive Backend service
- DeviceHive Auth service
- DeviceHive Admin Console
- DeviceHive Websocket Proxy
- DeviceHive Nginx Proxy
- Kafka
- PostgreSQL
- Hazelcast IMDG
- Install Docker Compose using official instructions.
git clone https://github.com/devicehive/devicehive-docker.git- cd rdbms-image
DeviceHive can be started without any additional configuration, docker-compose.yml file contains all necessary parameters set to safe defaults. Though there is one parameter that could and should be changed due to security reasons - JWT_SECRET.
DeviceHive ecosystem uses JWT tokens for authentication. JWT_SECRET is used for signing JWT tokens and is generated at startup of DeviceHive and stored in the database. You can change it by exporting the JWT_SECRET environment variable or by adding JWT_SECRET=<your value> line in the .env file inside current directory.
In order to run DeviceHive stack on top of Docker containers, define environment variables as per your requirements and run:
sudo docker-compose up -d
After a while (1-2 minutes) you should be able to access your DeviceHive micro services via endpoints described in the next section.
There's a huge amount of business cases which DeviceHive ecosystem can not cover from the scratch. For example, you may want to implement alerting in case of some emergency. Or to store sensor data outside, in AWS S3 or Cassandra. That's why Plugin management service were designed and developed. It allows to create your own simple consumers of notifications and commands, implement custom behavior with the Node.js, Python or Java templates and run as a separate process inside Docker container or as a Kubernetes pod.
To enable optional DeviceHive Plugin management service start docker-compose with the following command:
sudo docker-compose -f docker-compose.yml -f dh_plugin.yml up -d
Or add line COMPOSE_FILE=docker-compose.yml:dh_plugin.yml in .env file.
The DeviceHive MQTT plugin is MQTT transport layer between MQTT clients and DeviceHive server. The broker uses WebSocket sessions to communicate with DeviceHive Server and Redis server for persistence functionality.
To enable optional DeviceHive MQTT brokers run DeviceHive with the following command. This will start MQTT brokers on port 1883 and internal Redis container:
sudo docker-compose -f docker-compose.yml -f mqtt-brokers.yml
Or add line COMPOSE_FILE=docker-compose.yml:mqtt-brokers.yml in .env file.
This datasource was created to connect Grafana with DeviceHive to track commands and notifications by particular device. To enable optional Grafana service with DeviceHive datasource run DeviceHive with the following command:
sudo docker-compose -f docker-compose.yml -f grafana.yml up -d
Or add line COMPOSE_FILE=docker-compose.yml:grafana.yml in .env file.
This only adds Grafana container. After that you need to install plugin in it:
sudo docker-compose exec grafana bash -c 'grafana-cli plugins install devicehive-devicehive-datasource'
sudo docker-compose down
sudo docker-compose up
Grafana will be available at http://<devicehive-host>/grafana with default Grafana credentials.
Table below lists endpoints where you can find various DeviceHive services. Replace localhost with actual hostname of the docker daemon and start observing DeviceHive capabilities.
| Service | URL |
|---|---|
| Admin Console | http://localhost/admin |
| Frontend service API | http://localhost/api/rest |
| Auth service API | http://localhost/auth/rest |
| Plugin service API | http://localhost/plugin/rest |
| Frontend Swagger | http://localhost/api/swagger |
| Auth Swagger | http://localhost/auth/swagger |
| Plugin Swagger | http://localhost/plugin/swagger |
| Grafana | http://localhost/grafana |
| Plugin websocket proxy | ws://localhost/plugin/proxy |
| MQTT brocker | mqtt://localhost:1883 |
| Port | Service | Notes |
|---|---|---|
| 80,443 | Nginx proxy | Primary port for all services |
| 1883 | MQTT brokers | If enabled, see MQTT plugin section above |
| 2181 | Zookeeper | |
| 5432 | PostgreSQL | |
| 5701 | Hazelcast | |
| 7071 | Kafka metrics | If enabled, see Kafka metrics section below |
| 8080 | Frontend service | |
| 8090 | Auth service | |
| 8110 | Plugin service | If enabled, see Plugin management service section above |
| 9092 | Kafka | |
| 9395 | cAdvisor | If enabled, see cAdvisor metrics section below |
| 3000 | Grafana | If enabled, see Grafana datasource section above |
In order to run only DeviceHive 3d-party dependencies in Docker containers, simply run:
sudo docker-compose -f dev-proxy.yml up -d
sudo docker-compose -f dev-rpc.yml up -d
Then you'd be able to start all DeviceHive java services (backend, frontend, auth, plugin manager) by running java -jar devicehive-...-boot.jar
All containers are configured via environment variables and Docker Compose can pass variables from its environment to containers or read them from .env file. To make persistent configuration changes we will add parameters in the .env file in the current directory.
JWT_SECRET- changes the randomly generated JWT signing secret.
Released versions of devicehive-docker use stable DeviceHive images from DeviceHive Docker Hub repository. But if you want follow DeviceHive development add following parameters:
DH_TAG- tag for DeviceHive Frontend, Backend and Hazelcast images. Can be set todevelopmentto track development version of DeviceHive.DH_ADMIN_TAG- tag for DeviceHive Admin Console image. Can be set todevelopmentto track development version of Admin Console.
These variables are used by Frontend, Backend and PostgreSQL containers.
DH_POSTGRES_ADDRESS- Address of PostgreSQL server instance. Defaults topostgres, which is address of internal PostgreSQL container.DH_POSTGRES_PORT- Port of PostgreSQL server instance, defaults to5432if undefined.DH_POSTGRES_DB- PostgreSQL database name for DeviceHive metadata. It is assumed that database already exists and either blank or has been initialized by DeviceHive. Defaults topostgres.DH_POSTGRES_USERNAMEandDH_POSTGRES_PASSWORD- login/password for DeviceHive user in PostgreSQL that have full access toDH_POSTGRES_DB. Defaults arepostgresandmysecretpassword.
To enable DeviceHive to communicate over Apache Kafka message bus to scale out and interoperate with other componets, such us Apache Spark, or to enable support of Apache Cassandra for fast and scalable storage of device messages define the following environment variables:
DH_ZH_ADDRESS- Comma-separated list of addressed of ZooKeeper instances. Defaults tozookeeper, which is address of internal Zookeeper container.DH_ZK_PORT- Port of ZooKeeper instances, defaults to2181if undefined.DH_KAFKA_BOOTSTRAP_SERVERS- Comma separated list of Kafka servers, i.e.host1:9092,host2:9092,host3:9092. This parameter orDH_KAFKA_ADDRESSis required to be set.DH_KAFKA_ADDRESS- Address of Apache Kafka broker node. Mutually exclusive withDH_KAFKA_BOOTSTRAP_SERVERSparameter.DH_KAFKA_PORT- Port of Apache Kafka broker node, defaults to9092if undefined. Ignored ifDH_KAFKA_ADDRESSis undefined.DH_RPC_SERVER_REQ_CONS_THREADS- Kafka request consumer threads in the Backend, defaults to3if undefined.DH_RPC_SERVER_WORKER_THREADS- Server worker threads in the Backend, defaults to3if undefined. On machine with many CPU cores and high load this value must be raised. For example on machine with 8 core it must be set to6.DH_RPC_CLIENT_RES_CONS_THREADS- Kafka response consumer threads in the Frontend, defaults to3.DH_FE_SPRING_PROFILES_ACTIVE,DH_BE_SPRING_PROFILES_ACTIVEandDH_PLUGIN_SPRING_PROFILES_ACTIVE- Changes which Spring profile use for Frontend, Backend and Plugin sevices respectively. Defaults tows-kafka-proxy-frontendfor Frontend,ws-kafka-proxy-backendfor Backend andws-kafka-proxyfor Plugin. Can be changed torpc-clientfor Frontend/Plugin andrpc-serverfor Backend to use direct connection to Kafka instead of devicehive-ws-proxy service.
By default DeviceHive writes minimum logs for better performance. Two configuration parameters are supported:
DH_LOG_LEVEL- log verbosity for DeviceHive Java classes. Defaults toINFOfor both devicehive-frontend and devicehive-backend.ROOT_LOG_LEVEL- log verbosity for external dependencies. Defaults toWARNfor devicehive-frontend andINFOfor devicehive-backend.
Possible values are: TRACE, DEBUG, INFO, WARN, ERROR.
You can find more configurable parameters in frontend and backend startup scripts.
DeviceHive Proxy provides TLS support by default. If custom certificate is not configured it generates self-signed certificate and stores them in dh-proxy-ssl Docker volume.
For Docker Compose installation we will use Compose feature to read configuration from multiple Compose files. Second Compose file will start devicehive-proxy container with custom certificate.
To configure DeviceHive Proxy to use your own certificate follow next steps:
- Generate key and certificate signing request for your domain, sign CSR with Certificate Authority. Resulting certificate and key files must be in the PEM format.
- Create
ssldirectory outside ofdevicehive-dockerdirectory, on the same level. - Generate dhparam file for nginx:
openssl dhparam -out ssl/dhparam.pem 2048
- Copy SSL certificate to
ssldirectory. File must be namedssl_certificate. - Copy SSL certificate key to
ssldirectory. File must be namedssl_certificate_key. - Run DeviceHive with the following command:
sudo docker-compose -f docker-compose.yml -f dh_proxy_custom_certificate.yml up -d
Or add line COMPOSE_FILE=docker-compose.yml:dh_proxy_custom_certificate.yml in .env file.
You can now access your DeviceHive API at https://devicehive-host-url/api and Admin Console at https://devicehive-host-url/admin.
We provide Compose file with cAdvisor service which exports various container-related metrics. It's exposed on port 9395.
Run DeviceHive with the following command:
sudo docker-compose -f docker-compose.yml -f cadvisor.yml
Or add line COMPOSE_FILE=docker-compose.yml:cadvisor.yml in .env file.
You can start Kafka service with additional Prometheus metrics exporter. Necessary parameters for Kafka container are already configured in devicehive-metrics.yml file. It will launch JMX exporter on tcp port 7071.
Run DeviceHive with the following command:
sudo docker-compose -f docker-compose.yml -f devicehive-metrics.yml
Or add line COMPOSE_FILE=docker-compose.yml:devicehive-metrics.yml in .env file.
Related Prometheus config for this exporter and link to Grafana Dashboard is in the Monitoring Kafka with Prometheus blog post by Prometheus developer.
Continuous Integration system uploads images built from every branch to devicehiveci repository on Docker Hub.
To use these images add ci-images.yml to COMPOSE_FILE parameter in .env file. If you don't have this parameter in .env file, add it like that:
COMPOSE_FILE=docker-compose.yml:ci-images.yml
DeviceHive Frontend and Backend services can be run with remote JMX connection enabled. TCP ports 9999-10002 must be open on a firewall.
- Create
jmxremote.passwordandjmxremote.accessfile in the current directory.jmxremote.passwordmust be readable by owner only. For example, if you want to grant JMX access for user 'developer' with password 'devpass', create these files like that:
echo "developer devpass" > jmxremote.password
echo "developer readwrite" > jmxremote.access
chmod 0400 jmxremote.password
- Open
jmx-remote.ymlfile and replace<external hostname>in _JAVA_OPTIONS env vars with actual hostname of DeviceHive server. - Run DeviceHive with the following command:
sudo docker-compose -f docker-compose.yml -f jmx-remote.yml
Or add line COMPOSE_FILE=docker-compose.yml:jmx-remote.yml in .env file.
You can launch Management Center to monitor Hazelcast usage and health. TCP port 9980 must be open on a firewall.
- Add
hazelcast-management-center.ymltoCOMPOSE_FILEparameter in.envfile. If you don't have this parameter in.envfile, add it like that:
COMPOSE_FILE=docker-compose.yml:hazelcast-management-center.yml
- Run DeviceHive as usual.
- Open Hazelcast Management Center in browser via http://devicehive-server:9980/mancenter. You'll be required to configure authentication on the first launch.
To backup database use following command:
sudo docker-compose exec postgres sh -c 'pg_dump --no-owner -c -U ${POSTGRES_USER} ${POSTGRES_DB}' > dump_`date +%d-%m-%Y"_"%H_%M_%S`.sql
This will create dump_*.sql file in the current directory.
To restore database from SQL dump file delete existing database (if any), start only postgres container and pass dump file contents to psql utility in container:
sudo docker-compose down
sudo docker volume ls -q|grep devicehive-db| xargs sudo docker volume rm
sudo docker-compose up -d postgres
cat dump_*.sql | sudo docker exec -i rdbmsimage_postgres_1 sh -c 'psql -U ${POSTGRES_USER} ${POSTGRES_DB}'
sudo docker-compose up -d
Example configuration steps for CentOS 7.3 to became Docker host:
- Install CentOS 7.3, update it and reboot.
- Install docker-latest package:
sudo yum install -y docker-latest
-
Configure Docker to use LVM-direct storage backend. These steps are required for better disk IO performance:
- Add new disk with at least 10 GB of disk space. It will be used as physical volume for Docker volume group.
- Add following lines to
/etc/sysconfig/docker-latest-storage-setupfiles. Change/dev/xvdbfor you device.
VG=docker DEVS=/dev/xvdb- Run storage configuration utility
sudo docker-latest-storage-setup -
Enable and start Docker service:
sudo systemctl enable docker-latest
sudo systemctl start docker-latest
-
Install docker-compose:
- Install and update python-pip package manager:
sudo yum install -y python2-pip sudo pip install -U pip- Install docker-compose:
pip install docker-compose
Enjoy!