ekuiper-new/docs/en_US/edgex/edgex_rule_engine_tutorial.md

EdgeX rule engine tutorial

Overview

In EdgeX Geneva, EMQ X Kuiper - an SQL based rule engine is integrated with EdgeX. Before diving into this tutorial, let's spend a little time on learning basic knowledge of Kuiper. Kuiper is an edge lightweight IoT data analytics / streaming software implemented by Golang, and it can be run at all kinds of resource constrained edge devices. Kuiper rules are based on Source, SQL and Sink.

• Source: The data source of streaming data, such as data from MQTT broker. In EdgeX scenario, the data source is EdgeX message bus, which could be ZeroMQ or MQTT broker.
• SQL: SQL is where you specify the business logic of streaming data processing. Kuiper provides SQL-like statements to allow you to extract, filter & transform data.
• Sink: Sink is ued for sending analysis result to a specified target. For example, send analysis result to another MQTT broker, or an HTTP rest address.

Following three steps are required for using Kuiper.

• Create a stream, where you specify the data source.
• Write a rule.
  • Write a SQL for data analysis
  • Specify a sink target for saving analysis result
• Deploy and run rule.

The tutorial demonstrates how to use Kuiper to process the data from EdgeX message bus.

Kuiper EdgeX integration

EdgeX uses message bus to exchange information between different micro services. It contains the abstract message bus interface and an implementation for ZeroMQ & MQTT (NOTICE: ONLY ZeroMQ message bus is supported in Kuiper rule engine, MQTT will be supported in later versions). The integration work for Kuiper & EdgeX includes following 3 parts.

• An EdgeX message bus source is extended to support consuming data from EdgeX message bus.

• To analyze the data, Kuiper need to know data types that passed through it. Generally, user would be better to specify data schema for analysis data when a stream is created. Such as in below, a demo stream has a field named temperature field. It is very similar to create table schema in relational database system. After creating the stream definition, Kuiper can perform type checking during compilation or runtime, and invalid SQLs or data will be reported to user.

  CREATE STREAM demo (temperature bigint) WITH (FORMAT="JSON"...)

However, since data type definitions are already specified in EdgeX Core contract Service , and to improve the using experience, user are NOT necessary to specify data types when creating stream. Kuiper source tries to load all of value descriptors from Core contract Service during initialization of a rule (so now if you have any updated value descriptors, you will have to restart the rule), then if with any data sending from message bus, it will be converted into corresponding data types.

• An EdgeX message bus sink is extended to support send analysis result back to EdgeX Message Bus. User can also choose to send analysis result to RestAPI, Kuiper already supported it.

Start to use

In out tutorial, we will use Random Integer Device Service which is shipped in official released EdgeX, and run rules against the data generated by this sample device service.

Run EdgeX Docker instances

After the EdgeX Geneva is offcially released, you can just follow steps in this doc to start the service. But now since Kuiper has not been official released yet, you have to download Docker composer file from here, and then bring up EdgeX Docker instances.

# wget https://github.com/edgexfoundry/developer-scripts/raw/master/releases/nightly-build/compose-files/docker-compose-nexus-mongo-no-secty.yml

# docker-compose -f ./docker-compose-nexus-redis-no-secty.yml up -d --build

After all of the Docker instances are started, you can use docker ps command to verify all of services are runnings correctly.

TODO: The docker instance list

Create a stream

There are two approaches to manage stream, you can use your preferred approach.

Option 1: Use Rest API

The next step is to create a stream that can consume data from EdgeX message bus. Please change $your_server to Kuiper docker instance IP address.

curl -X POST \
  http://$TODO:9081/streams \
  -H 'Content-Type: application/json' \
  -d '{
  "sql": "create stream demo() WITH (FORMAT=\"JSON\", TYPE=\"edgex\")"
}'

For other Rest APIs, please refer to this doc.

Option 2: Use Kuiper CLI

Run following command to enter the running Kuiper docker instance.

docker exec -it kuiper /bin/sh

Use following command to create a stream named demo.

bin/cli create stream demo'() WITH (FORMAT="JSON", TYPE="edgex")'

For other command line tools, please refer to this doc.

---

Now the stream is created. But you maybe curious about how Kuiper knows the message bus IP address & port, because such information are not specified in CREATE STREAM statement. Those configurations are managed in etc/sources/edgex.yaml , you can type cat etc/sources/edgex.yaml command to take a look at the contents of file. If you have different server, ports & service server configurations, please update it accordingly. As mentioned previously, these configurations could be override when bring-up the Docker instances.

#Global Edgex configurations
default:
  protocol: tcp
  server: localhost
  port: 5563
  topic: events
  serviceServer: http://localhost:48080
.....  

For more detailed information of configuration file, please refer to this doc.

Create a rule

Let's create a rule that send result data to an MQTT broker, for detailed information of MQTT sink, please refer to this link. Similar to create a stream, you can also choose REST or CLI to manage rules.

So the below rule will filter all of randomnumber that is less than 31. The sink result will be published to topic result of public MQTT broker broker.emqx.io.

Option 1: Use Rest API

curl -X POST \
  http://$TODO:9081/rules \
  -H 'Content-Type: application/json' \
  -d '{
  "id": "rule1",
  "sql": "SELECT * FROM demo WHERE randomnumber > 30",
  "actions": [
    {
      "mqtt": {
        "server": "tcp://broker.emqx.io:1883",
        "topic": "result",
        "clientId": "demo_001"
      }
    }
  ]
}'

Option 2: Use Kuiper CLI

You can create a rule file with any text editor, and copy following contents into it. Let's say the file name is rule.txt.

{
  "sql": "SELECT * from demo",
  "actions": [
    {
      "mqtt": {
        "server": "tcp://broker.emqx.io:1883",
        "topic": "result",
        "clientId": "demo_001"
      }
    }
  ]
}

In the running Kuiper instance, and execute following command.

# bin/cli create rule rule1 -f rule.txt
Connecting to 127.0.0.1:20498...
Creating a new rule from file rule.txt.
Rule rule1 was created, please use 'cli getstatus rule $rule_name' command to get rule status.

---

If you want to send analysis result to another sink, please refer to other sinks that supported in Kuiper.

Now you can also take a look at the log file under log/stream.log, see detailed info of rule.

//TODO

Monitor analysis result

Since all of the analysis result are published to tcp://broker.emqx.io:1883, so you can just use below mosquitto_sub command to monitor the result. You can also use other MQTT client tools.

# mosquitto_sub -h broker.emqx.io -t result
//TODO

You'll find that only those randomnumber larger than 30 will be published to result topic.

You can also type below command to look at the rule execution status. The corresponding REST API is also available for getting rule status, please check related document.

# bin/cli getstatus rule rule1
Connecting to 127.0.0.1:20498...
{
  "source_demo_0_records_in_total": 29,
  "source_demo_0_records_out_total": 29,
  "source_demo_0_exceptions_total": 0,
  "source_demo_0_process_latency_ms": 0,
  "source_demo_0_buffer_length": 0,
  "source_demo_0_last_invocation": "2020-03-19T10:30:09.294337",
  "op_preprocessor_demo_0_records_in_total": 29,
  "op_preprocessor_demo_0_records_out_total": 29,
  "op_preprocessor_demo_0_exceptions_total": 0,
  "op_preprocessor_demo_0_process_latency_ms": 0,
  "op_preprocessor_demo_0_buffer_length": 0,
  "op_preprocessor_demo_0_last_invocation": "2020-03-19T10:30:09.294355",
  "op_filter_0_records_in_total": 29,
  "op_filter_0_records_out_total": 21,
  "op_filter_0_exceptions_total": 0,
  "op_filter_0_process_latency_ms": 0,
  "op_filter_0_buffer_length": 0,
  "op_filter_0_last_invocation": "2020-03-19T10:30:09.294362",
  "op_project_0_records_in_total": 21,
  "op_project_0_records_out_total": 21,
  "op_project_0_exceptions_total": 0,
  "op_project_0_process_latency_ms": 0,
  "op_project_0_buffer_length": 0,
  "op_project_0_last_invocation": "2020-03-19T10:30:09.294382",
  "sink_sink_mqtt_0_records_in_total": 21,
  "sink_sink_mqtt_0_records_out_total": 21,
  "sink_sink_mqtt_0_exceptions_total": 0,
  "sink_sink_mqtt_0_process_latency_ms": 0,
  "sink_sink_mqtt_0_buffer_length": 1,
  "sink_sink_mqtt_0_last_invocation": "2020-03-19T10:30:09.294423"
}

Summary

In this tutorial, we introduce a very simple use of EdgeX Kuiper rule engine. If having any issues regarding to use of Kuiper rule engine, you can open issues in EdgeX or Kuiper Github respository.

More Excecise

Current rule does not filter any data that are sent to Kuiper, so how to filter data? For example, if you only concern Int32 field value that are great than 30. Please drop rule and change the SQL in previous rule file as following.

{
  "sql": "SELECT * from demo WHERE Int32 > 30",
  "actions": [...]
}

After update the rule file, and then deploy the rule again. Please monitor the result topic of MQTT broker, you will find that only Int32 field value that are great than 30 will be sent out.

Extended Reading

• Read EdgeX source for more detailed information of configurations and data type conversion.
• How to use meta function to extract additional data from EdgeX message bus? There are some other information are sent along with device service, such as event created time, event id etc. If you want to use such metadata information in your SQL statements, please refer to this doc.
• EdgeX message bus sink doc. The document describes how to use EdgeX message bus sink. If you'd like to send the analysis result into message bus, you are probably interested in this article.

If you want to explore more features of EMQ X Kuiper, please refer to below resources.

• Kuiper Github code repository
• Kuiper reference guide