Skip to content

Flow Framework package to check the health status of you application

License

Notifications You must be signed in to change notification settings

t3n/flow-healthstatus

Repository files navigation

CircleCI Latest Stable Version Total Downloads License

t3n.Flow.HealthStatus

Package to check the health status of a flow application.

It's extremly useful in a kubernetes environment to use with readiness and liveness probe to determine if a pod can serve traffic and if it is still alive.

Usage

To determine the current health status of your application you can check wether the app is ready or still alive.

Readiness

Simply execute the flow command ./flow app:isready.

This will execute all tests defined in the t3n.Flow.HealthStatus.testChain of the Settings.yaml. If all tests have passed, the readyChain tasks will be executed.

After a successfully run of the readyChain an internal lock will be set to prevent repeated execution. The testChain will be executed on every run. So the readyChain should bring your application in an "ready state". Make sure to initialize everything you need. The testChain should ping all services your application depends on.

Liveness

Execute ./flow app:isalive to check if your pod is still alive.

This will execute the livenessChain.

Currently the liveness chain is empty by default and has one possible test: statusCode.

Configuration

Add all your tests in the following format in your apps Settings.yaml:

t3n:
  Flow:
    HealthStatus:
      testChain:
        yourUniqueTestKey:
          name: 'Optional name'
          test: 'database' // shorthand for a predefined task in t3n\Flow\HealthStatus\Test\*.Test or a full qualified class name
          options:
            key: 'value' // optional options for your test
          position: 'after otherTestKey' // optional position

After that, the check will execute the ready chain:

t3n:
  Flow:
    HealthStatus:
      readyChain:
        yourUniqueTaskKey:
          name: 'Optional name'
          task: 'command' // shorthand for a predefined task in t3n\Flow\HealthStatus\Task\*.Task or a full qualified class name
          options:
            key: 'value' // optional options for your task
          position: 'after otherTaskKey' / optional position
          lockName: 'mylock' // optional lock override. This will create a lock for this task only and ignore the global lock

After a successful ready chain invokation, you can call ./flow app:isalive to execute your liveness chain:

t3n:
  Flow:
    HealthStatus:
      livenessChain:
        yourUniqueTestKey:
          name: 'Optional name'
          task: 'statusCode' // shorthand for a predefined task in t3n\Flow\HealthStatus\LivenessTest\*.Test or a full qualified class name
          options:
            key: 'value' // optional options for your task
          position: 'after otherTaskKey' / optional position

Advanced configuration

Before each attempt to execute a ready task, the check will test the t3n.Flow.HealthStatus.defaultReadyTaskCondition to see if the task should be executed. In the default configuration this is simply a check to see if the ready lock is not yet set.

You can override this behaviour on a per task basis:

t3n:
  Flow:
    HealthStatus:
      readyChain:
        yourUniqueTaskKey:
          condition: '${Lock.isSet("mylock")}' // this can be any eel expression
          afterInvocation: '${Lock.set("mylock")}' // this will be executed after a successfull invocation

(the lockName setting is simply a shorthand for exactly this example)

To extend the eel context, you can provide additional helpers in t3n.Flow.HealthStatus.defaultContext.

Health status via HTTP-Request

If you'd like to check your Application via HTTP instead of a cli command you can do so by including the Routes in your Routes.yaml:

-
  name: 'Health-Routes'
  uriPattern: '<HealthStatusSubroutes>'
  subRoutes:
    'HealthStatusSubroutes':
      package: 't3n.Flow.HealthStatus'
      variables:
        'healthStatusEndpoint': 'your-endpoint-name'

Afterwards there will be two new routes: /<your-endpoint-name>/ready and /<your-endpoint-name>/live Adjust the variable to your needs. Both endpoints will return a JSON formatted output. On a successful run the response has a status code 200 and if there are any errors the status code will be 500.

Example Configuration

This example could be used in your Flow package to make sure that your application pod has a ready state to serve traffic. Therefore it will always check the ping status for doctrine, redis and beanstalk. On the first run all missing database migrations will be executed, the redis cache flushed and static resources published. After a successfull run only the testChain will be executed again.

t3n:
  Flow:
    HealthStatus:
      testChain:
        database:
          test: doctrine
          position: start
        redis:
          test: redis
          options:
            hostname: your-redis-host
        beanstalk:
          test: beanstalk
          options:
            hostname: your-beanstalk-host
      readyChain:
        migrations:
          task: command
          options:
            command: 'neos.flow:doctrine:migrate'
        flushRedis:
          name: 'Flush redis'
          position: 'start 100'
          task: redis
          options:
            hostname: your-redis-host
            command: FLUSHDB
            database: 0
        staticResources:
          name: 'Publish static resources'
          task: command
          position: 'end 20'
          lockname: staticresources
          cacheName: t3n_FlowHealthStatus_LocalLock
          options:
            command: 'neos.flow:resource:publish'
            arguments:
              collection: static
      livenessChain:
        home:
          test: statusCode
          name: 'Homepage responds'
          options:
            url: '/'
            method: 'GET'
            statusCode: 200

Note the lockname configuration. This Configuration enables you to run tasks only once per deployment or always. By default the t3n_FlowHealthStatus_Lock cache is used to read and write locks. Add this to your Caches.yaml and all your application pods will rely on the same lock files as they don't use the local file storage but redis. This will result in a execution once per deployment:

t3n_FlowHealthStatus_Lock:
  backend: Neos\Cache\Backend\RedisBackend
  backendOptions:
    hostname: 'your-redis-server'
    database: 2

The staticResources task has a custom cacheName configured. To ensure that this task will be executed in each application pod set it to local file storage:

t3n_FlowHealthStatus_LocalLock:
  frontend: Neos\Cache\Frontend\StringFrontend
  backend: Neos\Cache\Backend\FileBackend