Rule Engine Overview

• Key Concepts
  • Rule Engine Message
  • Rule Node
  • Rule Node Connection
  • Rule Chain
  • Message Processing Result
  • Rule Engine Queue
    • Queue submit strategy
    • Queue processing strategy
    • Default queues
• Predefined Message Types
• Rule Node Types
• Configuration
  • Test script functions
• Rule Engine Statistics
• Debugging
• Import/Export
• Architecture
• Custom REST API calls to Rule Engine
• Tutorials
• Troubleshooting

ThingsBoard Rule Engine is a highly customizable and configurable system for complex event processing. With rule engine you are able to filter, enrich and transform incoming messages originated by IoT devices and related assets. You are also able to trigger various actions, for example, notifications or communication with external systems.

Key Concepts

Rule Engine Message

Rule Engine Message is a serializable, immutable data structure that represent various messages in the system. For example:

• Incoming telemetry, attribute update or RPC call from device;
• Entity life-cycle event: created, updated, deleted, assigned, unassigned, attributes updated;
• Device status event: connected, disconnected, active, inactive, etc;
• Other system events.

Rule Engine Message contains the following information:

• Message ID: time based, universally unique identifier;
• Originator of the message: Device, Asset or other Entity identifier;
• Type of the message: “Post telemetry” or “Inactivity Event”, etc;
• Payload of the message: JSON body with actual message payload;
• Metadata: List of key-value pairs with additional data about the message.

Rule Node

Rule Node is a basic component of Rule Engine that process single incoming message at a time and produce one or more outgoing messages. Rule Node is a main logical unit of the Rule Engine. Rule Node can filter, enrich, transform incoming messages, perform action or communicate with external systems.

Rule Node Connection

Rule Nodes may be connected to other rule nodes. Each relation has relation type, a label used to identify logical meaning of the relation. When rule node produces the outgoing message it always specifies the relation type which is used to route message to next nodes.

Typical rule node relations are “Success” and “Failure”. Rule nodes that represent logical operations may use “True” or “False”. Some specific rule nodes may use completely different relation types, for example: “Post Telemetry”, “Attributes Updated”, “Entity Created”, etc.

Some rule nodes support custom connection names. Just type your custom connection name and click the “Create a new one!” link:

All connection names are case-sensitive.

Rule Chain

Rule Chain is a logical group of rule nodes and their relations. For example, the rule chain below will:

• save all telemetry messages to the database;
• raise “High Temperature Alarm” if temperature field in the message will be higher then 50 degrees;
• raise “Low Temperature Alarm” if temperature field in the message will be lower then -40 degrees;
• log failure to execute the temperature check scripts to console in case of logical or syntax error in the script.

Tenant administrator is able to define one Root Rule Chain and optionally multiple other rule chains. Root rule chain handles all incoming messages and may forward them to other rule chains for additional processing. Other rule chains may also forward messages to different rule chains.

For example, the rule chain below will:

• raise “High Temperature Alarm” if temperature field in the message will be higher then 50 degrees;
• clear “High Temperature Alarm” if temperature field in the message will be less then 50 degrees;
• forward events about “Created” and “Cleared” alarms to external rule chain that handles notifications to corresponding users.

Message Processing Result

There are three possible results of message processing: Success, Failure and Timeout. The message processing attempt is marked as “Success” when the last rule node in the processing chain successfully process the message. The message processing attempt is marked as “Failure” if one of the rule nodes produce “Failure” of message processing, and there is no rule nodes to handle that failure. The message processing attempt is marked as “Timeout” when overall time of processing exceed configurable threshold.

See diagram below and let’s review the possible scenarios:

If the “Transformation” script fails, the message is not marked as “Failed”, because there is a “Save to DB” node connected with “Failure” relation. If the “Transformation” script is successful, it will be pushed to “External System” with the REST API call. If the external system is overloaded, the REST API call may “hang” for some time. Let’s assume the overall timeout for message pack processing is 20 seconds. Let’s ignore Transformation script execution time because it is < 1ms. So, if the “External System” will reply within 20 seconds, the message will be successfully processed. Similar, if “Save to DB” call will succeed, the message will be successfully processed. However, if the external system will not reply within 20 seconds, the message processing attempt will be marked as “timed-out”. Similar, if “Save to DB” call will fail, the message will be marked as failed.

Rule Engine Queue

See new documentation

Queue submit strategy

See new documentation

Queue processing strategy

See new documentation

Default queues

See new documentation

Predefined Message Types

List of the predefined Message Types is presented in the following table:

Rule Node Types

All available rule nodes are grouped in correspondence with their nature:

• Filter Nodes are used for message filtering and routing;
• Enrichment Nodes are used to update meta-data of the incoming Message;
• Transformation Nodes are used for changing incoming Message fields like Originator, Type, Payload, Metadata;
• Action Nodes execute various actions based on incoming Message;
• External Nodes are used to interact with external systems.

Configuration

Each Rule Node may have specific configuration parameters that depend on the Rule Node Implementation. For example, “Filter - script” rule node is configurable via custom JS function that process incoming data. “External - send email” node configuration allows to specify mail server connection parameters.

Rule Node configuration window may be opened by double-clicking on the node in the Rule Chain editor:

Test script functions

Some rule nodes have specific UI feature that allow users to test TBEL/JS functions. Once you click on the Test Filter Function you will see the JS Editor that allows you to substitute input parameters and verify the output of the function.

You can define:

• Message Type in the top left field.
• Message payload in the left Message section.
• Metadata in right Metadata section.
• Actual TBEL/JS script in Filter section.

After pressing Test output will be returned in right Output section.

Rule Engine Statistics

ThingsBoard Team have prepared the “default” dashboard for Rule Engine statistics. This dashboard is automatically loaded for each tenant. You may access it by navigating to “Api Usage”->”View statistics” (see screen below).

The statistics collection is enabled by default and is controlled via configuration properties.

You may notice insights about errors in processing and what causes them on the dashbaord below:

Debugging

ThingsBoard provides ability to review incoming and outgoing messages for each Rule Node. To enable debug, user need to ensure that “Debug mode” checkbox is selected in the main configuration window (see first image in the Configuration section).

Once debug is enabled, user is able to see incoming and outgoing messages info as long as corresponding relation types. See image below for a sample debug messages view:

Import/Export

You are able to export your rule chain to JSON format and import it to the same or another ThingsBoard instance.

In order to export rule chain, you should navigate to the Rule Chains page and click on the export button located on the particular rule chain card.

Similar, to import the rule chain you should navigate to the Rules Chains page and click on the big “+” button in the bottom-right part of the screen and then click on the import button.

Architecture

To learn more about internals of the rule engine, see architecture page.

Custom REST API calls to Rule Engine

ThingsBoard PE Feature

Only Professional Edition supports Custom Rule Engine REST API calls feature.
Use ThingsBoard Cloud or install your own platform instance.

ThingsBoard provides API to send custom REST API calls to the rule engine, process the payload of the request and return result of the processing in response body. This is useful for a number of use cases. For example:

• extend existing REST API of the platform with custom API calls;
• enrich REST API call with the attributes of device/asset/customer and forward to external system for complex processing;
• provide custom API for your custom widgets.

To execute the REST API call, you may use rule-engine-controller REST APIs:

Note: the entity id you have specified in the call will be the originator of Rule Engine message. If you do not specify the entity id parameters, your user entity will become an originator of the message.

Tutorials

ThingsBoard authors have prepared several tutorials to help you get started with designing rule chains by example:

• Transform incoming messages from device
• Transform incoming messages using previous messages from device
• Create and clear alarms on incoming device messages
• Send emails to customer on device alarm
• Send messages between related devices

See more tutorials here.

Troubleshooting

If you are using Kafka queue for processing messages, ThingsBoard provides the ability to monitor if the rate of pushing messages to the Kafka is faster than rate of consuming and processing them (in such case you will have a growing latency for message processing). To enable this functionality, you need to ensure that Kafka consumer-stats are enabled (see queue.kafka.consumer-stats section of the Configuration properties)

Once Kafka consumer-stats are enabled, you will see logs (see Troubleshooting) about offset lag for consumer groups (there are consumer-group logs for tb-core, tb-rule-engine and all transport services).

Here’s an example of the log message:

1
2021-03-19 15:01:59,794 [kafka-consumer-stats-11-thread-1] INFO  o.t.s.q.k.TbKafkaConsumerStatsService - [re-Main-consumer] Topic partitions with lag: [[topic=[tb_rule_engine.main.0], partition=[0], committedOffset=[5413], endOffset=[5418], lag=[5]]].

From this message we can see that there are 5 (5418 - 5413 = 5) messages pushed to the Main queue (tb_rule_engine.main.0 Kafka topic) but not yet processed.

In general the logs have the following structure:

1
TIME [STATS_PRINTING_THREAD_NAME] INFO  o.t.s.q.k.TbKafkaConsumerStatsService - [CONSUMER_GROUP_NAME] Topic partitions with lag: [[topic=[KAFKA_TOPIC], partition=[KAFKA_TOPIC_PARTITION], committedOffset=[LAST_PROCESSED_MESSAGE_OFFSET], endOffset=[LAST_QUEUED_MESSAGE_OFFSET], lag=[LAG]],[topic=[ANOTHER_TOPIC], partition=[], committedOffset=[], endOffset=[], lag=[]],...].

Where:

• CONSUMER_GROUP_NAME - name of the consumer group which is processing messages (could be any of the rule-engine queues, core queue etc)
• KAFKA_TOPIC - name of the exact Kafka topic
• KAFKA_TOPIC_PARTITION - number of the topic’s partition
• LAST_PROCESSED_MESSAGE_OFFSET - the sequence number of the last message which was processed by the consumer (last acknowledged message in the Rule Engine etc)
• LAST_QUEUED_MESSAGE_OFFSET - the sequence number of the last message that was successfully pushed to the Kafka topic
• LAG - the amount of unprocessed messages

NOTE: Logs about consumer lag are printed only if there is a lag for this consumer group.