Public | Automated Build

Last pushed: 4 months ago
Short Description
kafka provider test builds
Full Description

OpenWhisk package for communication with Kafka or IBM Message Hub

This project is an OpenWhisk package that allows you to communicate with Kafka or IBM Message Hub instances for publishing and consuming messages using native high performance Kafka API.

OpenWhisk is a cloud-first distributed event-based programming service and represents an event-action platform that allows you to execute code in response to an event. These events can come from many different sources, such as Object Storage, direct HTTP, Cloudant database, etc. One of the popular event sources is Message Hub and Kafka, both of which can also be used as an effective instriment to publish events from OpenWhisk to the rest of the world and allow for highly scalable and performant asyncronous communication for event driven applications.

Using the Messaging package

This package allows you to create triggers that react when messages are posted to either an IBM Message Hub instance, or to a generic Kafka instance. Since the parameters required for each of these situations are different, there are two separate feeds to handle them: /messaging/messageHubFeed and messaging/kafkaFeed.

Additionally, two actions are included which allow you to produce messages to either Message Hub, or generic Kafka instances. These are /messaging/messageHubProduce, and /messaging/kafkaProduce respectively.

Creating a Trigger that listens to an IBM MessageHub instance

In order to create a trigger that reacts when messages are posted to a Message Hub instance, you need to use the feed named /messaging/messageHubFeed. This feed action supports the following parameters:

Name Type Description
kafka_brokers_sasl JSON Array of Strings This parameter is an array of <host>:<port> strings which comprise the brokers in your Message Hub instance
user String Your Message Hub user name
password String Your Message Hub password
topic String The topic you would like the trigger to listen to
kafka_admin_url URL String The URL of the Message Hub admin REST interface
isJSONData Boolean (Optional - default=false) When set to true this will cause the provider to attempt to parse the message value as JSON before passing it along as the trigger payload.
isBinaryKey Boolean (Optional - default=false) When set to true this will cause the provider to encode the key value as Base64 before passing it along as the trigger payload.
isBinaryValue Boolean (Optional - default=false) When set to true this will cause the provider to encode the message value as Base64 before passing it along as the trigger payload.

While this list of parameters may seem daunting, they can be automatically set for you by using the package refresh CLI command:

  1. Create an instance of Message Hub service under your current organization and space that you are using for OpenWhisk.

  2. Verify that the the topic you want to listen to already exists in Message Hub or create a new topic, for example mytopic.

  3. Refresh the packages in your namespace. The refresh automatically creates a package binding for the Message Hub service instance that you created.

    $ wsk package refresh
    
    created bindings:
    Bluemix_Message_Hub_Credentials-1
    
    $ wsk package list
    
    packages
    /myBluemixOrg_myBluemixSpace/Bluemix_Message_Hub_Credentials-1 private
    

    Your package binding now contains the credentials associated with your Message Hub instance.

  4. Now all you need to do is create a Trigger that will be fired when new messages are posted to your Message Hub topic.

    $ wsk trigger create MyMessageHubTrigger -f /myBluemixOrg_myBluemixSpace/Bluemix_Message_Hub_Credentials-1/messageHubFeed -p topic mytopic
    

Setting up a Message Hub package outside Bluemix

If you're not using OpenWhisk in Bluemix or if you want to set up your Message Hub outside of Bluemix, you must manually create a package binding for your Message Hub service. You need the Message Hub service credentials and connection information.

  1. Create a package binding that is configured for your Message Hub service.

    $ wsk package bind /whisk.system/messaging myMessageHub -p kafka_brokers_sasl "[\"kafka01-prod01.messagehub.services.us-south.bluemix.net:9093\", \"kafka02-prod01.messagehub.services.us-south.bluemix.net:9093\", \"kafka03-prod01.messagehub.services.us-south.bluemix.net:9093\"]" -p user <your Message Hub user> -p password <your Message Hub password> -p kafka_admin_url https://kafka-admin-prod01.messagehub.services.us-south.bluemix.net:443
    
  2. Now you can create a Trigger using your new package that will be fired when new messages are posted to your Message Hub topic.

    $ wsk trigger create MyMessageHubTrigger -f myMessageHub/messageHubFeed -p topic mytopic -p isJSONData true
    

Creating a Trigger that listens to a Generic Kafka instance

In order to create a trigger that reacts when messages are posted to an unauthenticated Kafka instance, you need to use the feed named messaging/kafkaFeed. This feed supports the following parameters:

Name Type Description
brokers JSON Array of Strings This parameter is an array of <host>:<port> strings which comprise the brokers in your Message Hub instance
topic String The topic you would like the trigger to listen to
isJSONData Boolean (Optional - default=false) When set to true this will cause the provider to attempt to parse the message value as JSON before passing it along as the trigger payload.
isBinaryKey Boolean (Optional - default=false) When set to true this will cause the provider to encode the key value as Base64 before passing it along as the trigger payload.
isBinaryValue Boolean (Optional - default=false) When set to true this will cause the provider to encode the message value as Base64 before passing it along as the trigger payload.

Example:

$ wsk trigger create MyKafkaTrigger -f /whisk.system/messaging/kafkaFeed -p brokers "[\"mykafkahost:9092\", \"mykafkahost:9093\"]" -p topic mytopic -p isJSONData true

Listening for messages

After creating a trigger, the system will monitor the specified topic in your messaging service. When new messages are posted, the trigger will be fired.

The payload of that trigger will contain a messages field which is an array of messages that have been posted since the last time your trigger fired. Each message object in the array will contain the following fields:

  • topic
  • partition
  • offset
  • key
  • value

In Kafka terms, these fields should be self-evident. However, key has an optional feature isBinaryKey that allows the key to transmit binary data. Additionally, the value requires special consideration. Optional fields isJSONData and isBinaryValue are available to handle JSON and binary messages. These fields, isJSONData and isBinaryValue, cannot be used in conjunction with each other.

As an example, if isBinaryKey was set to true when the trigger was created, the key will be encoded as a Base64 string when returned from they payload of a fired trigger.

For example, if a key of Some key is posted with isBinaryKey set to true, the trigger payload will resemble the below:

{
    "messages": [
        {
            "partition": 0,
            "key": "U29tZSBrZXk=",
            "offset": 421760,
            "topic": "mytopic",
            "value": "Some value"
        }
    ]
}

If the isJSONData parameter was set to false (or not set at all) when the trigger was created, the value field will be the raw value of the posted message. However, if isJSONData was set to true when the trigger was created, the system will attempt to parse this value as a JSON object, on a best-effort basis. If parsing is successful, then the value in the trigger payload will be the resulting JSON object.

For example, if a message of {"title": "Some string", "amount": 5, "isAwesome": true} is posted with isJSONData set to true, the trigger payload might look something like this:

{
  "messages": [
    {
      "partition": 0,
      "key": null,
      "offset": 421760,
      "topic": "mytopic",
      "value": {
          "amount": 5,
          "isAwesome": true,
          "title": "Some string"
      }
    }
  ]
}

However, if the same message content is posted with isJSONData set to false, the trigger payload would look like this:

{
  "messages": [
    {
      "partition": 0,
      "key": null,
      "offset": 421761,
      "topic": "mytopic",
      "value": "{\"title\": \"Some string\", \"amount\": 5, \"isAwesome\": true}"
    }
  ]
}

Similar to isJSONData, if isBinaryValue was set to true during trigger creation, the resultant value in the trigger payload will be encoded as a Base64 string.

For example, if a value of Some data is posted with isBinaryValue set to true, the trigger payload might look something like this:

{
  "messages": [
    {
      "partition": 0,
      "key": null,
      "offset": 421760,
      "topic": "mytopic",
      "value": "U29tZSBkYXRh"
    }
  ]
}

If the same message is posted without isBinaryData set to true, the trigger payload would resemble the below example:

{
  "messages": [
    {
      "partition": 0,
      "key": null,
      "offset": 421760,
      "topic": "mytopic",
      "value": "Some data"
    }
  ]
}

Messages are batched

You will notice that the trigger payload contains an array of messages. This means that if you are producing messages to your messaging system very quickly, the feed will attempt to batch up the posted messages into a single firing of your trigger. This allows the messages to be posted to your trigger more rapidly and efficiently.

Please keep in mind when coding actions that are fired by your trigger, that the number of messages in the payload is technically unbounded, but will always be greater than 0. Here is an example of a batched message (please note the change in the offset value):

 {
   "messages": [
       {
         "partition": 0,
         "key": null,
         "offset": 100,
         "topic": "mytopic",
         "value": {
             "amount": 5
         }
       },
       {
         "partition": 0,
         "key": null,
         "offset": 101,
         "topic": "mytopic",
         "value": {
             "amount": 1
         }
       },
       {
         "partition": 0,
         "key": null,
         "offset": 102,
         "topic": "mytopic",
         "value": {
             "amount": 999
         }
       }
   ]
 }

Producing messages to Message Hub

If you would like to use an OpenWhisk action to conveniently produce a message to Message Hub, you can use the /messaging/messageHubProduce action. This action takes the following parameters:

Name Type Description
kafka_brokers_sasl JSON Array of Strings This parameter is an array of <host>:<port> strings which comprise the brokers in your Message Hub instance
user String Your Message Hub user name
password String Your Message Hub password
topic String The topic you would like the trigger to listen to
value String The value for the message you would like to produce
key String (Optional) The key for the message you would like to produce
base64DecodeValue Boolean (Optional - default=false) If true, the message will be produced with a Base64 decoded version of the value parameter
base64DecodeKey Boolean (Optional - default=false) If true, the message will be produced with a Base64 decoded version of the key parameter

While the first three parameters can be automatically bound by using wsk package refresh, here is an example of invoking the action with all required parameters:

wsk action invoke /messaging/messageHubProduce -p kafka_brokers_sasl "[\"kafka01-prod01.messagehub.services.us-south.bluemix.net:9093\", \"kafka02-prod01.messagehub.services.us-south.bluemix.net:9093\", \"kafka03-prod01.messagehub.services.us-south.bluemix.net:9093\"]" -p topic mytopic -p user <your Message Hub user> -p password <your Message Hub password> -p value "This is the content of my message"

Producing messages to a generic Kafka instance

If you would like to use an OpenWhisk action to conveniently produce a message to a generic Kafka instance, you can use the /messaging/kafkaProduce action. This action takes the following parameters:

Name Type Description
brokers JSON Array of Strings This parameter is an array of <host>:<port> strings which comprise the brokers in your Kafka cluster
topic String The topic you would like the trigger to listen to
value String The value for the message you would like to produce
key String (Optional) The key for the message you would like to produce
base64DecodeValue Boolean (Optional - default=false) If true, the message will be produced with a Base64 decoded version of the value parameter
base64DecodeKey Boolean (Optional - default=false) If true, the message will be produced with a Base64 decoded version of the key parameter

Here is an example of invoking the action with all required parameters:

wsk action invoke /messaging/kafkaProduce -p brokers "[\"mykafkahost:9092\", \"mykafkahost:9093\"]" -p topic mytopic -p value "This is the content of my message"

Producing Messages with Binary Content

You may find that you want to use one of the above actions to produce a message that has a key and/or value that is binary data. The problem is that invoking an OpenWhisk action inherently involves a REST call to the OpenWhisk server, which may require any binary parameter values of the action invocation to be Base64 encoded. How to handle this?

The action caller (you, or your code) must first Base64 encode the data, for example, the value of the message you want to produce. Pass this encoded data as the value parameter to the produce action. However, to ensure that the produced message's value contains the original bytes, you must also set the base64DecodeValue parameter to true. This will cause the produce action to first Base64 decode the value parameter before producing the message. The same procedure applies to producing messages with a binary key, using the base64DecodeKey parameter set to true in conjunction with a Base64 encoded key parameter.

Examples

Integrating OpenWhisk with IBM Message Hub, Node Red, IBM Watson IoT, IBM Object Storage, and IBM Data Science Experience

Example that integrates OpenWhisk with IBM Message Hub, Node Red, IBM Watson IoT, IBM Object Storage, IBM Data Science Experience (Spark) service can be found here.

Testing

An OpenWhisk deployment is required in order for the automated test suite to be executed. To run tests locally, run $ ./gradlew tests:test -Dhost=<container_address> -Dport=<docker_port> from the project's root directory. Where <docker_address> is the IP address of the service's docker container, and <docker_port> is the port the docker container is listening on. In addition, OPENWHISK_HOME must be set to the root of the local OpenWhisk directory. Ex: export OPENWHISK_HOME=<openwhisk_directory>.

References

Docker Pull Command
Owner
jasonpet

Comments (0)