This section describes the specific South-bound API implemented by this IoTAgent. For the Configuration API and other APIs concerning general IoTAgents, check the API Reference section;
Ultralight 2.0 is a lightweight text based protocol aimed to constrained devices and communications where the bandwidth and device memory may be limited resources.
The payload for information update requests is composed of a list of key-value pairs separated by the | character.
E.g.:
t|15|k|abc
In this example, two attributes, one named "t" with value "15" and another named "k" with value "abc" are transmitted. Values in Ultralight 2.0 are not typed (everything is treated as a string).
Multiple groups of measures can be combined into a single request (but just for HTTP/POST or MQTT), using the #
character. In that case, a different NGSI request will be generated for each group of measures. E.g.:
gps|1.2/3.4#t|10
This will generate two elements in the NGSI batch update request (POST /v2/op/update) for the same entity, one for each one of the measures. Each one of those elements can contain any number of attributes.
Measure groups can additionally have an optional timestamp, with the following syntax:
2016-06-13T00:35:30Z|lle|100
The timestamp will be added as a prefix of the measures themselves, separated by a '|'. The attribute will be translated
to a TimeInstant attribute in the final entity.T
Current version of the agent only supports active attributes, i.e. those attributes actively reported +by the device to the agent. Passive or lazy attributes, i.e. those attributes whose value is only given upon explicit +request from the agent, are not implemented. Please check the issue +#23 for more details and updates regarding its implementation.
Commands are messages sent to the device from the IoT Agent. A command has the following format:
<device name>@<command name>|<command value>
This indicates that the device (named 'device_name' in the Context Broker) has to execute the command 'command_name', with the given value. E.g.:
Robot1@turn|left
This example will tell the Robot 1 to turn to left.
In the case of complex commands requiring parameters, the command_value could be used to implement parameter passing.
E.g:
weatherStation167@ping|param1=1|param2=2
This example will tell the Weather Station 167 to reply to a ping message with the provided params.
Once the command has finished its execution in the device, the reply to the server must adhere to the following format:
<device name>@<command name>|result
Where device_name and command_name must be the same ones used in the command execution, and the result is the final
result of the command. E.g.:
weatherStation167@ping|Ping ok
In this case, the Weather station replies with a String value indicating everything has worked fine.
The latest versions of the Provisioning API allow for the definition of reverse expressions to keep data shared between the Context Broker and the device in sync (regardless of whether the data originated in plain data from the device or in a transformation expression in the IoTAgent). In this cases, when a reverse expression is defined, whenever the bidirectional attribute is modified, the IoTAgent sends a command to the original device, with the name defined in the reverse expression attribute and the ID of the device (see Commands Syntax, just above).
It is possible to use expressions to transform commands, in the same way that other attributes could do it, that is
adding expression to command definition. This way a command could be defined like:
{
"name": "reset",
"type": "command",
"expression": "{ set: 0}"
}and when command will be executed the command value will be the result of apply value to defined expression. Following the example case the command will be:
set|0
Additionally a command could define a payloadType in their definition with the aim to transform payload command with
the following meanings:
- binaryfromstring: Payload will transformed into a be Buffer after read it from a string.
- binaryfromhex: Payload will transformed into a be Buffer after read it from a string hex.
- binaryfromjson: Payload will transformed into a be Buffer after read it from a JSON string.
- json: Payload will be stringify from a JSON.
<empty>: This is the default case. Payload will not be transformed.
FIXME: this need to be tested, once IOTA Lib 3.0.0 gets released and IOTA UL 2.0.0 (using it) gets released.
Ultralight 2.0 defines a method that allows to use native JSON types in the NGSI v2. For example: The IotAgent receives this UL measure:
t|10|s|true|l|78.8
then the NGSI v2 update uses 10(number), true (boolean) and 78.8 (number) instead of "10" (string), "true"
(string) and "78.8" (string).
This functionality relies on string measures casting feature implemented in the iotagent library. This functionality
uses native JavaScript
JSON.parse() function
to cast data coming from measures (as text) to JSON native types. This functionality does not change the attribute type,
using the type specified in the config group or device provision, even if it is not consistent with the measures that
are coming. As an example, for a given measure:
a|1|b|1.01|c|true|d|null|e|[1,2,3]|f|['a','b','c']|g|{a:1,b:2,c:3}|h|I'm a string
The resulting entity would be something like:
{
"id": "entityid:001",
"type": "entitytype",
"a": {
"type": "provisionedType",
"value": 1
},
"b": {
"type": "provisionedType",
"value": 1.01
},
"c": {
"type": "provisionedType",
"value": true
},
"d": {
"type": "provisionedType",
"value": null
},
"e": {
"type": "provisionedType",
"value": [1, 2, 3]
},
"f": {
"type": "provisionedType",
"value": ["a", "b", "c"]
},
"g": {
"type": "provisionedType",
"value": { "a": 1, "b": 2, "c": 3 }
},
"h": {
"type": "provisionedType",
"value": "I'm a string"
}
}Note that provisionedType is the type included in the device provision or config group, and it is not changed.
Ultralight 2.0 defines a payload describing measures and commands to share between devices and servers but, does not specify a single transport protocol. Instead, different transport protocol bindings can be established for different scenarios.
The following sections describe the bindings currently supported: HTTP, MQTT and AMQP.
There are three possible interactions defined in the HTTP binding: requests with GET, requests with POST and commands.
A device can report new measures to the IoT Platform using an HTTP GET request to the /iot/d path with the following
query parameters:
- i (device ID): Device ID (unique for the API Key).
- k (API Key): API Key for the service the device is registered on.
- t (timestamp): Timestamp of the measure. Will override the automatic IoTAgent timestamp (optional).
- d (Data): Ultralight 2.0 payload.
Payloads for GET requests should not contain multiple measure groups.
Another way of reporting measures is to do it using a POST request. In this case, the payload is passed along as the request payload. Two query parameters are still mandatory:
- i (device ID): Device ID (unique for the API Key).
- k (API Key): API Key for the service the device is registered on.
- t (timestamp): Timestamp of the measure. Will override the automatic IoTAgent timestamp (optional).
All the interations between IotAgent and ContextBroker related to comamnds are described in Theory: Scenario 3: commands and Practice: Scenario 3: commands - happy path and Practice: Scenario 3: commands - error.
MQTT devices commands are always push. For HTTP Devices commands to be push they must be provisioned with the
endpoint attribute, from device or group device, that will contain the URL where the IoT Agent will send the received
commands. Otherwise the command will be poll. When using the HTTP transport, the command handling have two flavours:
- Push commands: The request payload format will be the one described in the UL Protocol description. The device will reply with a 200OK response containing the result of the command in the UL2.0 result format. Example of the HTTP request sent by IOTA in the case of push command:
POST http://[DEVICE_IP]:[PORT]
fiware-service: smart
fiware-servicepath: /streetligths
content-type: text/plain
Robot1@turn|left
- Polling commands: in this case, the Agent does not send any messages to the device, being the later responsible of retrieving them from the IoTAgent whenever the device is ready to get commands. In order to retrieve commands from the IoT Agent, the device will send the query parameter 'getCmd' with value '1' as part of a normal measure. As a result of this action, the IoTAgent, instead of returning an empty body (the typical response to a measurement report), will return a list of all the commands available for the device, sepparated by the character '#'. The command payload is described in the commands syntax section (and its shared with the push commands). Whenever the device has completed the execution of the command, it will send the response in the same way measurements are reported, but using the command result format as exposed in the commands syntax section.
Some additional remarks regarding polling commands:
- Commands can be also retrieved without needed of sending a mesaure. In other words, the device is not forced to send
a measure in order to get the accumulated commands. However, in this case note that
GETmethod is used to carry thegetCmd=1query parameter (as they are no actual payload for measures,POSTwouldn't make too much sense). Example to retrieve commands from IoT Agent:
curl -X GET 'http://localhost:7896/iot/d?i=motion001&k=4jggokgpepnvsb2uv4s40d59ov&getCmd=1' -i
- Example of the HTTP response sent by IOTA in the case of polling commands (and only one command is stored for that device):
200 OK
Content-type: text/plain
Robot1@turn|left
HTTP bindig is returning in a HTTP header named X-Processing-Time processing time (in milliseconds) expended by current HTTP measure
request. For example:
X-Processing-Time: 38
MQTT is a machine-to-machine (M2M)/IoT connectivity protocol, focused on a lightweight interaction between peers. MQTT is based on publish-subscribe mechanisms over a hierarchical set of topics defined by the user.
This section specifies the topics and messages allowed when using MQTT as the transport protocol for Ultralight 2.0. All the topics subscribed by the agent (to send measures, to configuration command retrieval or to get result of a command) are prefixed with the agent procotol:
/ul/<apiKey>/<deviceId>
where <apiKey> is the API Key assigned to the service and <deviceId> is the ID of the device.
All topics published by the agent (to send a comamnd or to send configuration information) to a device are not prefixed
by the protocol, in this case '/ul', just include apikey and deviceid (e.g: /FF957A98/MydeviceId/cmd and
/FF957A98/MyDeviceId/configuration/values ).
Note Measures and commands are sent over different MQTT topics:
- Measures are sent on the
/<protocol>/<api-key>/<device-id>/attrstopic,- Commands are sent on the
/<api-key>/<device-id>/cmdtopic,The reasoning behind this is that when sending measures northbound from device to IoT Agent, it is necessary to explicitly identify which IoT Agent is needed to parse the data. This is done by prefixing the relevant MQTT topic with a protocol, otherwise there is no way to define which agent is processing the measure. This mechanism allows smart systems to connect different devices to different IoT Agents according to need.
For southbound commands, this distinction is unnecessary since the correct IoT Agent has already registered itself for the command during the device provisioning step and the device will always receive commands in an appropriate format.
This transport protocol binding is still under development.
In order to send a single measure value to the server, the device must publish the plain value to the following topic:
/ul/<apiKey>/<deviceId>/attrs/<attrName>
Where <apiKey> and <deviceId> have the typical meaning and <attrName> is the name of the measure the device is
sending.
or instance, if using Mosquitto with a device with ID id_sen1, API Key ABCDEF and
attribute IDs h and t, then humidity measures are reported this way:
$ mosquitto_pub -t /ul/ABCDEF/id_sen1/attrs/h -m 70 -h <mosquitto_broker> -p <mosquitto_port> -u <user> -P <password>In order to send multiple measures in a single message, a device must publish a message in the following topic:
/ul/<apiKey>/<deviceId>/attrs
Where <apiKey> and <deviceId> have the typical meaning. The payload of such message should be a legal Ultralight 2.0
payload (with or without measure groups).
For instance, if using Mosquitto with a device with ID id_sen1, API Key ABCDEF and
attribute IDs h and t, then all measures (humidity and temperature) are reported this way:
$ mosquitto_pub -t /ul/ABCDEF/id_sen1/attrs -m 'h|70|t|15' -h <mosquitto_broker> -p <mosquitto_port> -u <user> -P <password>The protocol offers a mechanism for the devices to retrieve its configuration (or any other value it needs from those
stored in the Context Broker). Two topics are created in order to support this feature: a topic for configuration
commands and a topic to receive configuration information. This mechanism can be enabled or disabled using a
configuration flag, configRetrieval.
In case of MQTT to retrieve configuration parameters from the Context Broker, it is required that the device should be provisioned using "MQTT" as transport key, at device or group level. By default it will be considered "HTTP" as transport.
The parameter will be given as follows:
"transport": "MQTT"
This mechanism and the bidirectionality plugin cannot be simultaneously activated.
/ul/{{apikey}}/{{deviceid}}/configuration/commands
The IoT Agent listens in this topic for requests coming from the device. The messages must contain an Ultralight 2.0 payload with the following format:
{{type}}|{{fields}}
- type: indicates the type of command the device is sending. See below for accepted values.
- fields: array with the names of the values to be retrieved from the Context Broker entity representing the
device, separated by the
|character.
This command will trigger a query to the CB that will, as a result, end up with a new message posted to the Configuration information topic (described bellow).
E.g.:
configuration|pollingInterval|publishInterval
There are two accepted values for the configuration command types:
subscription: this command will generate a subscription in the Context Broker that will be triggered whenever any of the selected values change. In case the value has changed, all the attributes will be retrieved.configuration: this commands will generate a single request to the Context Broker from the IoTAgent, that will trigger a single publish message in the values topic.
/{{apikey}}/{{deviceid}}/configuration/values
Every device must subscribe to this topic, so it can receive configuration information. Whenever the device requests any information from the IoTA, the information will be posted in this topic. The information is published in the same format used in multiple command reporting: a plain Ultralight 2.0 text with:
- the
device idandcommand typeseparated by the@character; - a
|character; - a list of
attribute=valuerequested pairs separated by the|character.
An additional parameter called dt is added with the system current time.
E.g.:
device_1@configuration|pollingInterval=200|publishInterval=80|dt=20190626T154200Z
All the interations between IotAgent and ContextBroker related to comamnds are described in Theory: Scenario 3: commands and Practice: Scenario 3: commands - happy path and Practice: Scenario 3: commands - error.
Commands using the MQTT transport protocol binding always work in PUSH mode: the server publishes a message in a topic where the device is subscribed: the commands topic. Once the device has finished with the command, it publishes it result to another topic.
The commands topic, where the client will be subscribed has the following format:
/<apiKey>/<deviceId>/cmd
The result of the command must be reported in the following topic:
/ul/<apiKey>/<deviceId>/cmdexe
The command execution and command reporting payload format is specified under the Ultralight 2.0 Commands Syntax, above.
For instance, if a user wants to send a command ping with parameters data = 22, he will send the following request
to the Context Broker regarding an entity called sen1 of type sensor:
{
"updateAction": "UPDATE",
"contextElements": [
{
"id": "sen1",
"type": "sensor",
"isPattern": "false",
"attributes": [
{
"name": "ping",
"type": "command",
"value": "22"
}
]
}
]
}If the API key associated to de device is ABCDEF, and the device ID related to sen1 entity is id_sen1, this will
generate a message in the /ABCDEF/id_sen1/cmd topic with the following payload:
id_sen1@ping|22
If using Mosquitto, such a command is received by running the mosquitto_sub script:
$ mosquitto_sub -v -t /# -h <mosquitto_broker> -p <mosquitto_port> -u <user> -P <password> /ABCDEF/id_sen1/cmd id_sen1@ping|22At this point, Context Broker will have updated the value of ping_status to PENDING for sen1 entity. Neither
ping_info nor ping are updated.
Once the device has executed the command, it can publish its results in the /ul/ABCDEF/id_sen1/cmdexe topic with a
payload with the following format:
id_sen1@ping|1234567890
If using Mosquitto, such command result is sent by running the mosquitto_pub script:
$ mosquitto_pub -t /ul/ABCDEF/id_sen1/cmdexe -m 'id_sen1@ping|1234567890' -h <mosquitto_broker> -p <mosquitto_port> -u <user> -P <password>In the end, Context Broker will have updated the values of ping_info and ping_status to 1234567890 and OK,
respectively. ping attribute is never updated.
Some additional remarks regarding MQTT commands:
- MQTT devices can configure (at provisioning and updating time) each command with different values of MQTT QoS and MQTT retain values, which will be used only by a command. Moreover, in the same MQTT device different commands can be configured to use different MQTT options related with QoS level and Retain message policy. I.E:
{
"commands": [
{
"type": "command",
"name": "a_command_name_A",
"mqtt": { "qos": 2, "retain": true }
},
{
"type": "command",
"name": "a_command_name_B",
"mqtt": { "qos": 1, "retain": false }
}
]
}AMQP stands for Advance Message Queuing Protocol, and is one of the most popular protocols for message-queue systems. Although the protocol itself is software independent and allows for a great architectural flexibility, this transport binding has been designed to work with the RabbitMQ broker, in a way that closely resembles the MQTT binding (in the previous section). In fact, for IoT Platform deployments in need of an scalable MQTT Broker, RabbitMQ with the MQTT plugin will be used, connecting the IoT Agent to RabbitMQ through AMQP and the clients to RabbitMQ through MQTT.
The binding connects the IoT Agent to an exchange (usually amq.topic) and creates two queues (to share between all the
instances of the IoTAgents in a cluster environment): one for the incoming measures, and another for command result
update messages (named as the measure one, adding the _commands sufix).
For both measure reporting and command update status the mechanism is much the same as in the case of the MQTT binding: all the messages must be published to the selected exchange, using the following routing keys:
| Key pattern | Meaning |
|---|---|
| ...attrs | Multiple measure reporting |
| ...attrs. | Single measure reporting |
| ...cmd | Command reception |
| ...cmdexe | Command update message |
The payload is the same as for the other bindings.
The Ultralight 2.0 IoT Agent can work with multiple different transports for the same Ultralight 2.0 payload. Those
transports are dinamically loaded when the Agent starts, by looking in the lib/bindings folder for Node.js Modules.
Those module must export the following fields:
-
deviceProvisioningHandler(device, callback): this handler will be called each time a new device is provisioned in the IoT Agent. The device object contains all the information provided in the device registration.
-
configurationHandler(configuration, callback): handler for changes (provisioning or updates) in device groups. This handler should be used when configuration groups require any initialization or registration in the protocol binding.
-
start(newConfig, callback): starts the binding module, with the provided configuration. The
newConfigobject contains the global Agent configuration; the module should use a specific attribute inside the global scope to hold all its configuration values instead of using the global configuration scope itself. -
stop(callback): stops the binding module.
-
protocol: This field must contain a string key identifying the protocol. Requests coming from the server (commands and passive attributes) will use the
protocolfield of the devices and the correspondingprotocolattribute in the modules to identify which module should attend the request.
All the methods must call the callback before exiting (with or without error). Bindings will use methods in the IoT Agent Node.js library to interact process incoming requests.
The project is managed using npm.
For a list of available task, type
npm runThe following sections show the available options in detail.
Runs a local version of the IoT Agent
# Use git-bash on Windows
npm startMocha Test Runner + Should.js Assertion Library.
The test environment is preconfigured to run BDD testing style.
Module mocking during testing can be done with proxyquire
To run tests, type
docker run -d -p 27017:27017 mongo:4.2
docker run -d -p 5672:5672 rabbitmq:3.8.9
docker run -d -p 1883:1883 eclipse-mosquitto:1.6.7
npm testESLint
Uses the provided .eslintrc.json flag file. To check source code style, type
npm run lintSupport for continuous testing by modifying a src file or a test. For continuous testing, type
npm run test:watchIf you want to continuously check also source code style, use instead:
npm run watchIstanbul
Analizes the code coverage of your tests.
To generate an HTML coverage report under site/coverage/ and to print out a summary, type
# Use git-bash on Windows
npm run test:coverageremark
To check consistency of the Markdown markup, type
npm run lint:mdtextlint
Uses the provided .textlintrc flag file. To check for spelling and grammar errors, dead links and keyword consistency,
type
npm run lint:textRemoves node_modules and coverage folders, and package-lock.json file so that a fresh copy of the project is
restored.
# Use git-bash on Windows
npm run cleanRuns the prettier code formatter to ensure consistent code style (whitespacing, parameter placement and breakup of long lines etc.) within the codebase.
# Use git-bash on Windows
npm run prettierTo ensure consistent Markdown formatting run the following:
# Use git-bash on Windows
npm run prettier:text