What are Microservices?

A microservice in PIPEFORCE is an application which runs inside a (Docker) container which in term runs inside the PIPEFORCE (Kubernetes) cluster. Such a microservice can communicate with other microservices inside the same namespace via messaging and REST. A namespace can “host” many such microservices:

PIPEFORCE offers a lot of toolings to develop, deploy and monitor your microservices in your namesace. This is called the Microservices Framework.

Microservice or App?

In general, you have three main possibilities to write a business application:

• Using low code inside an app.

• Using a microservice.

• Combining both approaches.

Depending on the requirements you should select the right approach for your implementation.

In the table below you can see the pros and cons:

It's also possible to combine apps and microservices in order to implement a single business solution:

• The app should contain:

  • Connections to external systems

  • Data mappings and transformations

  • Forms and lists

  • Stateful workflows

• The microservice should contain:

  • Complex business specific logic

Designing a Microservice​

Typically a microservice is a relatively small application which has the responsibility about a concrete and well-defined part of an overall business process. How you slice the microservices depends on your requirements.

So before you start, you should be very clear, which parts should go into which microservices and which not at all.

See here for a good introduction how to design microservices: https://martinfowler.com/articles/break-monolith-into-microservices.html

Develop a Microservice​

Developing a microservice typically means, developing a business application in the programming language of your choice.

As long as you can build and run the application inside a (Docker) container, you can also run it inside PIPEFORCE as microservice.

We suggest you to write your microservices in one of these languages since they are widely used, having a huge community and being well documented:

1. Python. For example using one of the official Docker images: https://hub.docker.com/_/python

2. Java. For example using one of the frameworks Spring, Quarkus or Helidon and one of the official Docker images: https://hub.docker.com/_/java

3. NodeJS. For example using Typescript, ExpressJS and/or NestJS and one of the offcial Docker images: https://hub.docker.com/_/node

In our experience, Python is often a very good choice for a microservice language. Since it has the best trade-off between complexity and flexibility. But as always, it depends on your concrete requirements.

If you would like to start a microservice in Python, you can fork our template from GitHub and start coding it: https://github.com/logabit/pipeforce-service-template-python

In order to allow your microservice to communicate with others, there are two common ways you can implement this:

• Sync communication - Typically used with RESTful services inside PIPEFORCE.

• Async communication - Typically used with RabbitMQ and messaging inside PIPEFORCE (preferred way).

Deploy a Microservice​

Once a microservice has been developed, it must be wrapped inside a (Docker) image in order to be able to deploy it into PIPEFORCE. The deployment cycle of a microservice in PIPEFORCE is always an 4-step task:

1. Build a (Docker) container image from the sources of your microservice.

2. Upload the container image to a container registry which is supported by PIPEFORCE.

3. Deploy the image from this registry into your PIPEFORCE namespace by using the command service.start.

Step 2-4 are typically automated by using a CI/CD tool like Jenkins, CircleCI, Travis or similar.

Furthermore, we suggest managing your source code using GitHub and connect your CI/CD tool, so it starts to build, test and deploy automatically every time, a new push to GitHub happened (or on other triggers like merging or tagging).

Also see these commands to manage the deployment of your services:

• service.start - Starts a new service by deploying an image.

• service.stop - Stops a running service.

• service.status - Returns the status of a service.

Implicit ENV variables​

Every time a service is started using the service.start command, also some implicit variables will be automatically passed to this container and can be accessed via the environment variables inside the container.

These implicit variables are:

• PIPEFORCE_DOMAIN = The domain name used for the PIPEFORCE instance. For example customer.pipeforce.net.

• PIPEFORCE_HUB_URL = The cluster internal url of the hub service.

• PIPEFORCE_NAMESPACE = The namespace of the instance this services runs inside.

• PIPEFORCE_SERVICE = The name of this custom service inside PIPEFORCE.

• PIPEFORCE_MESSAGING_HOST = The messaging host to connect to in order to register a RabbitMQ listener.

• PIPEFORCE_MESSAGING_PORT = The messaging host to connect to in order to register a RabbitMQ listener.

• PIPEFORCE_MESSAGING_DEFAULT_DLQ = The default dead letter queue used for RabbitMQ messaging.

• PIPEFORCE_MESSAGING_DEFAULT_TOPIC = The default messaging topic used for RabbitMQ messaging.

Additionally to these default environment variables, you can also set your custom ones by using the parameter env on the command service.start:

pipeline:
  - service.start:
      name: myservice
      image: myimage
      env:
        MY_ENV: "myCustomValue"

Passing secrets as ENV​

In case you would like to set secret values to environment variables, you should create such secrets in the secret store inside PIPEFORCE and refer to them, using the custom uri prefix $uri:secret:. For example:

pipeline:
  - service.start:
      name: myservice
      image: myimage
      env:
        MY_SECRET_ENV: "$uri:secret:MY_SECRET"

On startup of the service, the secret with name MY_SECRET will be read from the secret store and passed to the container as environment variable MY_SECRET_ENV. This way it is not required to store the secret in code.

You can also define a default value for a secret. In case the secret could not be found in the secret store, the given default value will be used and passed as ENV variable value instead:

$uri:secret:MY_SECRET:someDefaultValue

Monitoring a microservice​

Logging

Everything you log into the standard output of your microservice container can be later viewed by using the command log.list and specifying the name under which you have deployed the service. Example:

pipeline:
  - log.list:
      service: "my-service"
      lines: "100"

Additionally you can use the log listing in the web portal to filter and search the logs of your microservice:

Service Status

The Services view in the web portal will show you the deployment status of your services. Here you can install new services, see their status and stop running services:

Also see these commands to manage your services:

• service.start - Starts a new service by deploying an image.

• service.stop - Stops a running service.

• service.status - Returns the status of a service.

Messaging in a Microservice

The preferred way each microservice inside PIPEFORCE can communicate with each other is by using messaging.

As message broker RabbitMQ is used by default.

If you're using one of our microservice templates, setting up a library to connect to RabbitMQ and creating the default channels and bindings is typically already done for you, or there is a best practise to do so.

Otherwise follow the documentation on the official RabbitMQ websites.

If you're interested in how to send and receive messages using pipelines, see Messaging and Events Framework.

How to connect a Microservice with the message broker?​

In case you have deployed your microservice using the command service.start, then these environment variables will be automatically provided inside your microservice and can be used to connect to the RabbitMQ message broker:

• PIPEFORCE_MESSAGING_HOST = The cluster-internal hostname of the messaging service.

• PIPEFORCE_MESSAGING_PORT = The cluster-internal port of the messaging service.

Additionally to these variables, you need to set the PIPEFORCE_MESSAGING_USERNAME and PIPEFORCE_MESSAGING_PASSWORD along with your service.start command using the custom uri prefix $uri:secret. Here is how to do it:

1. Create a new secret in your secret store with name pipeforce-messaging-username and type secret-text and set the RabbitMQ username you would like to use to connect.

2. Create a new secret in your secret store with name pipeforce-messaging-password and type secret-text and set the RabbitMQ password you would like to use to connect.

Start your container using these env variables pointing to the secret store. Example:

pipeline:
    - service.start:
        name: "myservice"
        image: "some/image"
        env:
            PIPEFORCE_MESSAGING_USERNAME: "$uri:secret:pipeforce-messaging-username"
            PIPEFORCE_MESSAGING_PASSWORD: "$uri:secret:pipeforce-messaging-password"

This way, the sensitive values will be passed to your container without the requirement to store them into code or refer from external systems.

In case you're using one of the custom service templates from PIPEFORCE, messaging will be automatically setup for you using the ENV variables mentioned before.

In case you're using your own framework, how to setup your client in order to connect, depends on your programming language the RabbitMQ client. See the official documentation: https://www.rabbitmq.com/devtools.html

The next step is to create the queues required by your service. See the Default Queue Naming guide below.

Now create a binding between your queues, the default topic pipeforce.hub.default.topic and the routing key pattern of messages your service is interested in.

Also remember to setup the dead leader queue in order to not lose any message as mentioned below.

Default Queue Naming​

By default any microservice is responsible to setup and manage its own queues.

Each queue should contain always equal message types. So, for different messages, create additional queues.

Queues created by a microservice should follow this naming convention:

<serviceName>_<queueName>_q

Whereas <serviceName> is the name of the microservice and <queueName> can be any name of the queue. For example:

shoppingcart_orders_q

Do not use dashes - but replace them by underscores _. For example instead of:

service-shoppingcart_orders_q

Use this name:

service_shoppingcart_orders_q

Default Topic​

PIEPFORCE automatically creates a default topic exchange on startup with this name: pipeforce.hub.default.topic.

PIPEFORCE core services are configured in a way that any event which happens there or is sent using the event.send command is also send to this default topic.

In case a microservice wants to listen to a certain type of message with a given routing key, it needs to create a binding between the topic pipeforce.hub.default.topic and the queue you want to “feed” this message into.

See here for more details about topics, routings and queues: https://www.rabbitmq.com/tutorials/tutorial-five-python.html

Default Dead Letter Queue​

Additionally, a default Dead Letter Queue is automatically configured by PIPEFORCE: pipeforce_hub_default_dlq.

Any other queue can be configured in a way to forward messages to this queue if one these rules apply:

• The message has been expired (was not processed within a given time period).

• The message has been declined (nack) by a consumer.

• The maximum amount of messages has been reached for a given queue.

This default dead letter queue is configured in a way to lazy keep the messages and store them into the persistence layer, so no messages get lost.

Typically in order to setup a dead letter queue for your custom queue, you have to set these arguments accordingly:

x-dead-letter-exchange = ""
x-dead-letter-routing-key = "pipeforce_hub_default_dlq"

How to set these arguments in your microservice depends on your selected programming language and the RabbitMQ client implementation you're using. See documentation for details: https://www.rabbitmq.com/dlx.html

In case you're using Java + Spring as microservice language for example, it could look like this:

return QueueBuilder.durable(Constants.MY_QUEUE)
    .withArgument("x-dead-letter-exchange", "")
    .withArgument("x-dead-letter-routing-key", "pipeforce_hub_default_dlq)
    .build();

Default Message Keys​

Here you can find the default message keys, PIPEFORCE will use for internal events and send messages with these keys to the default topic pipeforce.hub.default.topic. You can subscribe to these keys using a binding between your queue and this default topic.

You can also describe to multiple keys using the * and # patterns. For example in order to listen to all property events, you could subscribe to the key pattern property.*.

See here for more details about message key patterns on topics: https://www.rabbitmq.com/tutorials/tutorial-five-python.html