Status

Current state: [ UNDER DISCUSSION | ACCEPTED | REJECTED ]IMPLEMENTED

Discussion thread: <link to mailing list DISCUSS thread>

JIRA: SAMZA-TBD

Released: 

Problem

:http://mail-archives.apache.org/mod_mbox/samza-dev/201807.mbox/%3CCAFvExu3_nmaSQTy=5SypzwmqGA7S9+Txa=QkyERQ+hT3JZ29ig@mail.gmail.com%3E

JIRA:

Released: 

Problem

In the current implementation of ApplicationRunner, there In the current implementation of ApplicationRunner, there are a few issues: 

1. Instantiation of specific implementation of ApplicationRunner is exposed to the user, which requires user to choose a specific implementation of ApplicationRunner in source code, depending on the deployment environment (I.e. YARN vs standalone). 

2. ApplicationRunner only supports high-level API and does not fully support low-level API: 

   1. In standalone environment, user's program written in StreamTask/AsyncStreamTask classes is only supported in LocalApplicationRunner w/ a run() method 

   2. In YARN, RemoteApplicationRunner only support high-level API applications and falls back to JobRunner for low-level API applications. 

3. There is no unified API to allow user to specify initialization logic in high-level API and low-level API in initialization either. 

4. There is no defined standard lifecycle of a user application process in both YARN and standalone deployment. Hence, no consistent pattern to insert user code into the application’s full lifecycle 

   1. There is no standard method to insert user-defined application initialization sequence 

      1. In YARN, all application processes are initialized (i.e. configure re-writer, stream processor initialization, etc.) by the build-in main functions in Samza framework (i.e. ApplicationRunnerMain on launch host and LocalContainerRunner on NodeManagers). 

      2. In standalone, user can put arbitrary code in user main function to initialize the application process. 

   2. There is no defined method to allow user to inject customized logic before start/stop the user defined processors (I.e. StreamProcessors defined by user application) either, in addition to initialization 

...

1. Hide the choice of specific implementation of ApplicationRunner via configuration, not in source code. 

2. Define a unified API to allow user to specify describe the processing logic in high- and low-level API in all environment (I.e. all ApplicationRunners) w/ all lifecycle methods  

3. Expand the ApplicationRunner to run both lowDefine a set of standard application execution methods to deploy both low- and high-level APIs applications  in YARN and standalone environments. 

4. Define a standard processor life-cycle aware API to allow user’s users to inject customized logic to be injected before and after start/stop the processors in both YARN and standalone environments. 

Note that we need to define the following concepts clearly: 

1. Application’s execution methods in runtime are the functions to ApplicationRunner defines a set of standard execution methods to change the deployment status of an application in runtime instance (I.e. run/status/kill/waitForFinish) 
3. ApplicationRunner’s execution methods are the deployment environment specific implementation to the above API methods 
Application’s processor lifecycle aware methods are the user-defined functions to inject customized logic to be executed before or after we start or stop the stream processing logic defined by in the user application (I.e. beforeStart/afterStart/beforeStop/afterStop are called when we start/stop the StreamProcessors in local host) 

...

The proposed changes are the followings: 

1. Define a unified API ApplicationBaseAPI SamzaApplication as a single entry point for users to implement all user-customized logic, for both high-level API and low-level API 

   1. User implements a single describe() method to implement all user processing logic before creating the runtime application instance 

      1. Sub-classes StreamApplication and TaskApplication provide specific describe() methods for high-level API and low-level API, respectively 

2. Define a unified API class ApplicationSpecApplicationDescriptor to contain 

   1. High- and low-level processing logic defined via ApplicationBaseSamzaApplication.describe() . Sub-class StreamAppDescriptor and TaskAppDescriptor are used for high- and low-level APIs respectively.

   2. User implemented ProcessorLifecycleListenerFactory interface that creates a ProcessorLifecycleListener which includes customized logic User implemented ProcessorLifecycleListener class that includes customized logic to be invoked before and after starting/stopping the StreamProcessor(s) in the user application 

      1. Methods are beforeStart/afterStart/beforeStop/afterStop 

   3. Other used-defined objects in an application (e.g. configuration and context) (NOTE: this would be updated by SEP-15)

3. Expand ApplicationRunner with a mandatory constructor with SamzaApplication and Configure object as parameter

   1. An ApplicationRunner is now constructed with a user implemented SamzaApplication and Configure as the parameters

      1. SamzaApplication and Configure defines all user customized logic and configuration for an application 

      2. ApplicationRunner deploys

   Define the class ApplicationRuntime to represent a runtime instance of a user application, with a set of standard execution methods to change the deployment status of the runtime application instance in different environments (I.e. YARN and standalone)

   1. A runtime instance of user application is defined as a bundle of both 

      1. ApplicationSpec that contains all user customized logic. 

      2. ApplicationRunner that deploys and runs the user code. This would be instantiated from the configuration, not exposed to user at all. 

   2. The set of standard execution methods for an ApplicationRuntime object includes run/kill/status/waitForFinish 

A high-level overview of the proposed A high-level overview of the proposed changes is illustrated below: 
Image Removed

Image Added

Figure-1: high-level user programming model 

Image RemovedImage Added

Figure-2: Interaction and lifecycle of runtime API objects (using StreamApplication in LocalApplicationRunner as an example). 

...

The above design achieves the following goals: 

1. Defined a unified ApplicationBaseunified SamzaApplication interface for both high- and low-level APIs in different deployment environments. All user code is now implemented in one of the sub-classes (I.e. StreamApplicaiton or TaskApplication).  

   1. All processing logic is implemented in the standard describe() method in either StreamApplication or TaskApplication. 

   2. All user customized logic to start/stop contextual objects in their application process are in standard lifecycle listener methods (I.e. beforeStart/afterStart/beforeStop/afterStop)defined in ProcessorLifecycleListener. 

2. Construction of ApplicationRuntimeApplicationRunner object is implemented by Samza framework code, which hides: 

   1. Choice of a specific ApplicationRunner for different environment via configuration 

   2. Association of a user application to the specific instance of ApplicationRunner  as the parameter to the constructor

   3. Initialization of user processing logic before ApplicationRunner executes the application 

   4. Invoking user-defined lifecycle listener methods when run/kill the application via ApplicationRunner in local process (I.e. LocalApplicationRunner*) 

      1. Note that RemoteApplicationRunner only submit the application w/o launching the StreamProcessors. Hence, lifecycle listener methods are not invoked in RemoteApplicationRunner. 

      2. Note this is also pending on one refactor item that we need to refactor LocalContainerRunner s.t. 

         1. It implements run/kill/status w/ proper async implementation 

         2. It launches StreamProcessor instead of directly running SamzaContainer 

...

For Samza system build-in main method (as in ApplicationRunnerMain#main()), we require the user application class to have a default constructor w/o any parameters: 

Class<ApplicationBase>Class<SamzaApplication> appClass = (Class<ApplicationBase>Class<SamzaApplication>) Class.forName(appConfig.getAppClass());
if (StreamApplication.class.isAssignableFrom(appClass) || TaskApplication.class.isAssignableFrom(appClass)) {
  return appClass.newInstance();
}

The reason is: when deploying via RemoteApplicationRunner , in the NodeManagerYARN, we  we will run the managed main() method implemented by SamzaSamza in the NodeManager, which don’t have the ability to invoke customized constructor for user application. Hence, we expect a default constructor implemented by any user application. This is the same behavior as we expected today from any user implementing a high- or low-level application.  

...

1. The user application class implements a default constructor w/o any parameters 

2. Creation of ApplicationRuntimeApplicationRunner in main is using Samza provided methods (I.e. ApplicationRuntimesApplicationRunners.getApplicationRuntimegetApplicationRunner()) 

Simple code examples of high- and low-level API applications: 

public class PageViewCounterExample implements StreamApplication {

  // local execution mode
  publicpublic static void main(String[] args) {
    CommandLine cmdLine = new CommandLine();
    Config config = cmdLine.loadConfig(cmdLine.parser().parse(args));
    ApplicationRuntimeApplicationRunner appRuntimerunner = ApplicationRuntimesApplicationRunners.getApplicationRuntimegetApplicationRunner(ApplicationClassUtils.fromConfig(config), config);
    appRuntimerunner.startrun();
    appRuntimerunner.waitForFinish();
  }

  @Override
  public void describe(StreamApplicationSpecStreamApplicationDescriptor graphappDesc) {
      // Detailed examples on  MessageStream<PageViewEvent> pageViewEvents = null;how to create InputDescriptor and OutputDescriptor are in SEP-14
      pageViewEventsInputDescriptor pveStreamDescriptor = graph.getInputStreamgetInputDescriptor("pageViewEventStream", new JsonSerdeV2<>(PageViewEvent.class));
      OutputStream<KV<String,OutputDescriptor PageViewCount>>outputDescriptor pageViewEventPerMemberStream =
          graph.getOutputStream( getOutputDescriptor("pageViewEventPerMemberStream",
              KVSerde.of(new StringSerde(), new JsonSerdeV2<>(PageViewCount.class));
      MessageStream<PageViewEvent> pageViewEvents = null;
      pageViewEvents = appDesc.getInputStream(pveStreamDescriptor);
      OutputStream<KV<String, PageViewCount>> pageViewEventPerMemberStream =
          appDesc.getOutputStream(outputDescriptor);

      SupplierFunction<Integer> initialValue = () -> 0;
      FoldLeftFunction<PageViewEvent, Integer> foldLeftFn = (m, c) -> c + 1;
      pageViewEvents
          .window(Windows.keyedTumblingWindow(m -> m.memberId, Duration.ofSeconds(10), initialValue, foldLeftFn,
              null, null)
              .setEarlyTrigger(Triggers.repeat(Triggers.count(5)))
              .setAccumulationMode(AccumulationMode.DISCARDING), "tumblingWindow")
          .map(windowPane -> KV.of(windowPane.getKey().getKey(), new PageViewCount(windowPane)))
          .sendTo(pageViewEventPerMemberStream);
  }
}

...

public class TaskApplicationExample implements TaskApplication {

  public static void main(String[] args) {
    CommandLine cmdLine = new CommandLine();
    Config config = cmdLine.loadConfig(cmdLine.parser().parse(args));
    ApplicationRuntimeApplicationRunner appRuntimerunner = ApplicationRuntimes.getApplicationRuntimegetApplicationRunner(new TaskApplicationExample(), config);
    appRuntimerunner.startrun();
    appRuntimerunner.waitForFinish();
  }

  @Override
  public void describe(TaskApplicationSpecTaskAppDescriptor appBuilderappDesc) {
    // add input and output streams Detailed examples on how to create InputDescriptor and OutputDescriptor are in SEP-14
    appBuilder.addInputStreams(Collections.singletonListInputDescriptor pveStreamDescriptor = getInputDescriptor("myinput"));
    appBuilder.addOutputStreams(Collections.singletonList("myoutput"pageViewEventStream", new JsonSerdeV2<>(PageViewEvent.class));
    TableDescriptorOutputDescriptor tdoutputDescriptor = new RocksDbTableDescriptorgetOutputDescriptor("mytable")pageViewEventPerMemberStream",
        KVSerde.of(new StringSerde(), new JsonSerdeV2<>(PageViewCount.class));
    appBuilderappDesc.addTables(Collections.singletonList(td))addInputStream(pveStreamDescriptor);
     // create the task factory based on configurationappDesc.addOutputStream(outputDescriptor);
    TableDescriptor td = new RocksDbTableDescriptor("mytable");
    appBuilderappDesc.setTaskFactory(TaskFactoryUtiladdTable(td);
    // create the task factory based on configuration
    appDesc.setTaskFactory(TaskFactoryUtil.createTaskFactory(appBuilderappDesc.getConfig()));
  }

}

Public Interfaces

...

A) user-implemented interface classes include the followings: 

ApplicationBaseSamzaApplication: defines the basic life-cycle aware methods to allow users to inject customized logic before and after the lifecycle methods of an application 

public interface ApplicationBase<TSamzaApplication<S extends ApplicationSpec>ApplicationDescriptor> { 
  void describe(TS appSpecappDesc); 
} 

StreamApplication: extends ApplicationBaseSamzaApplication with a typed describe() method for high-level user application 

public interface StreamApplication extends ApplicationBase<StreamApplicationSpec>SamzaApplication<StreamApplicationDescriptor> { 
}

TaskApplication: extends ApplicationBasextends SamzaApplication with a typed describe() method to initialize the low-level user application 

public interface TaskApplication extends ApplicationBase<TaskApplicationSpec>SamzaApplication<TaskApplicationDescriptor> { 
}

ProcessorLifecycleListenerProcessorLifecycleListenerFactory: defines the unified processor lifecycle aware methods to allow users to inject customized logic before/after start/stop the StreamProcessor(s) in an application 

public interface ProcessorLifecycleListener { 
  default void beforeStart() {} 
  default void afterStart() {} 
  default void beforeStop() {} 
  default void afterStop() {} 
}

B) Samza framework implemented runtime objects 

Samza framework generates two sets of runtime classes that are directly exposed to the user. One set of classes are ApplicationSpec that includes all user-defined logic and configuration for an application; the other set of classes are runnable application instances that are a bundle of the ApplicationSpec and the corresponding ApplicationRunner. 

ApplicationSpec classes  

ApplicationSpec: this is a base interface for both high- and low-level applications. 

public interface ApplicationSpec<T extends ApplicationBase> { 
  /** 
   * Get the global unique application ID in the runtime process 
   * @return globally unique application ID 
   */ 
  String getGlobalAppId(); 
 
  /** 
   * Get the user defined {@link Config} 
   * @return config object 
   */ 
  Config getConfig(); 
 
  /** 
   * Sets the {@link ContextManager} for this {@link ApplicationSpec}. 
   * <p> 
   * The provided {@link ContextManager} can be used to setup shared context between the operator functions 
   * within a task instance 
   * 
   * @param contextManager the {@link ContextManager} to use for the {@link StreamApplicationSpec} 
   * @return the {@link StreamApplicationSpec} with {@code contextManager} set as its {@link ContextManager} 
   */ 
  ApplicationSpec<T> withContextManager(ContextManager contextManager); 


  /** 
   * Sets the lifecycle listener for user customized logic before and after starting/stopping 
   * StreamProcessors in the application 
   */  
  ApplicationSpec<T> withProcessorLifecycleListener(ProcessorLifecycleListener listener); 
} 

StreamApplicationSpec: this extends ApplicationSpec for a high-level application, including all methods to describe a high-level application in StreamGraph. 

public interface StreamApplicationSpec extends ApplicationSpec<StreamApplication>, StreamGraph { 
} 

...

factory interface to create ProcessorLifecycleListener in an application

public interface ProcessorLifecycleListenerFactory extends Serializable {
  /**
   * Create an instance of {@link ProcessorLifecycleListener} for the StreamProcessor
   *
   * @param pContext the context of the corresponding StreamProcessor
   * @param config the configuration of the corresponding StreamProcessor
   * @return the {@link ProcessorLifecycleListener} callback object for the StreamProcessor
   */
  ProcessorLifecycleListener createInstance(ProcessorContext pContext, Config config);
}


/**
 * The context for a StreamProcessor. This is a stub class, just include the method to identify the current StreamProcessor.
 *
 */
public interface ProcessorContext extends Serializable {
  String getProcessorId();
}

ProcessorLifecycleListener: defines the unified processor lifecycle aware methods to allow users to inject customized logic before/after start/stop the StreamProcessor(s) in an application 

public interface ProcessorLifecycleListener {
  /**
   * User defined initialization before a StreamProcessor is started
   */
  default void beforeStart() {}

  /**
   * User defined callback after a StreamProcessor is started
   *
   */
  default void afterStart() {}

  /**
   * User defined callback after a StreamProcessor is stopped successfully
   */
  default void afterStop() {}

  /**
   * User defined callback after a StreamProcessor is stopped with failure
   *
   * @param t the error causing the stop of the StreamProcessor.
   */
  default void afterFailure(Throwable t) {}
}

B) Samza framework implemented runtime objects 

Samza framework generates two sets of runtime classes that are directly exposed to the user. One set of classes are ApplicationDescriptorImpl that includes all user-defined logic and configuration for an application; the other set of classes are ApplicationRunner class. 

ApplicationDescriptor and ApplicationDescriptorImpl classes  

ApplicationDescriptor: this is a base interface for both high- and low-level applications. The corresponding implementation class is ApplicationDescriptorImpl.

public interface ApplicationDescriptor<S extends ApplicationDescriptor> { 
  /**
   * Get the {@link Config} of the application
   * @return config of the application
   */
  Config getConfig();

  /**
   * Sets the {@link ContextManager} for this application.
   * <p>
   * Setting the {@link ContextManager} is optional. The provided {@link ContextManager} can be used to build the shared
   * context between the operator functions within a task instance
   *
   * TODO: this should be replaced by the shared context factory when SAMZA-1714 is fixed.

   * @param contextManager the {@link ContextManager} to use for the application
   * @return type {@code S} of {@link ApplicationDescriptor} with {@code contextManager} set as its {@link ContextManager}
   */
  S withContextManager(ContextManager contextManager);

  /**
   * Sets the {@link ProcessorLifecycleListenerFactory} for this application.
   *
   * <p>Setting a {@link ProcessorLifecycleListenerFactory} is optional to a user application. It allows users to
   * plug in optional code to be invoked in different stages before/after the main processing logic is started/stopped in
   * the application.
   *
   * @param listenerFactory the user implemented {@link ProcessorLifecycleListenerFactory} that creates lifecycle listener
   *                        with callback methods before and after the start/stop of each StreamProcessor in the application
   * @return type {@code S} of {@link ApplicationDescriptor} with {@code listenerFactory} set as its {@link ProcessorLifecycleListenerFactory}
   */
  S withProcessorLifecycleListenerFactory(ProcessorLifecycleListenerFactory listenerFactory);

  /**
   * Sets a set of customized {@link MetricsReporterFactory}s in the application
   *
   * @param reporterFactories the map of customized {@link MetricsReporterFactory}s to be used
   * @return type {@code S} of {@link ApplicationDescriptor} with {@code reporterFactories}
   */
  S withMetricsReporterFactories(Map<String, MetricsReporterFactory> reporterFactories);
} 

StreamApplicationDescriptor: this extends ApplicationDescriptor for a high-level application, including all methods to describe a high-level application in a graph. The corresponding implementation is StreamApplicationDescriptorImpl.

public interface StreamApplicationDescriptor extends ApplicationDescriptor<StreamApplicationDescriptor> { 
  /**
   * Sets the default SystemDescriptor to use for intermediate streams. This is equivalent to setting
   * {@code job.default.system} and its properties in configuration.
   * <p>
   * If the default system descriptor is set, it must be set <b>before</b> creating any input/output/intermediate streams.
   * <p>
   * If an input/output stream is created with a stream-level Serde, they will be used, else the serde specified
   * for the {@code job.default.system} in configuration will be used.
   * <p>
   * Providing an incompatible message type for the intermediate streams that use the default serde will result in
   * {@link ClassCastException}s at runtime.
   *
   * @param defaultSystemDescriptor the default system descriptor to use
   * @return type {@code S} of {@link ApplicationDescriptor} with {@code defaultSystemDescriptor} set as its default system
   */
  StreamApplicationDescriptor withDefaultSystem(SystemDescriptor<?> defaultSystemDescriptor);

  /**
   * Gets the input {@link MessageStream} corresponding to the {@code inputDescriptor}.
   * <p>
   * A {@code MessageStream<KV<K, V>}, obtained by calling this method with a descriptor with a {@code KVSerde<K, V>},
   * can receive messages of type {@code KV<K, V>}. An input {@code MessageStream<M>}, obtained using a descriptor with
   * any other {@code Serde<M>}, can receive messages of type M - the key in the incoming message is ignored.
   * <p>
   * A {@code KVSerde<NoOpSerde, NoOpSerde>} or {@code NoOpSerde} may be used for the descriptor if the
   * {@code SystemConsumer} deserializes the incoming messages itself, and no further deserialization is required from  
   * the framework.
   * <p>
   * Multiple invocations of this method with the same {@code inputDescriptor} will throw an
   * {@link IllegalStateException}.
   *
   * @param inputDescriptor the descriptor for the stream
   * @param <M> the type of messages in the input {@link MessageStream}
   * @return the input {@link MessageStream}
   * @throws IllegalStateException when invoked multiple times with the same {@code inputDescriptor}
   */
  <M> MessageStream<M> getInputStream(InputDescriptor<M, ?> inputDescriptor);

  /**
   * Gets the {@link OutputStream} corresponding to the {@code outputDescriptor}.
   * <p>
   * An {@code OutputStream<KV<K, V>>}, obtained by calling this method with a descriptor with a {@code KVSerde<K, V>},
   * can send messages of type {@code KV<K, V>}. An {@code OutputStream<M>}, obtained using a descriptor with any
   * other {@code Serde<M>}, can send messages of type M without a key.
   * <p>
   * A {@code KVSerde<NoOpSerde, NoOpSerde>} or {@code NoOpSerde} may be used for the descriptor if the
   * {@code SystemProducer} serializes the outgoing messages itself, and no prior serialization is required from
   * the framework.
   * <p>
   * When sending messages to an {@code OutputStream<KV<K, V>>}, messages are partitioned using their serialized key.
   * When sending messages to any other {@code OutputStream<M>}, messages are partitioned using a null partition key.
   * <p>
   * Multiple invocations of this method with the same {@code outputDescriptor} will throw an
   * {@link IllegalStateException}.
   *
   * @param outputDescriptor the descriptor for the stream
   * @param <M> the type of messages in the {@link OutputStream}
   * @return the {@link OutputStream}
   * @throws IllegalStateException when invoked multiple times with the same {@code outputDescriptor}
   */
  <M> OutputStream<M> getOutputStream(OutputDescriptor<M, ?> outputDescriptor);

  /**
   * Gets the {@link Table} corresponding to the {@link TableDescriptor}.
   * <p>
   * Multiple invocations of this method with the same {@link TableDescriptor} will throw an
   * {@link IllegalStateException}.
   *
   * @param tableDescriptor the {@link TableDescriptor}
   * @param <K> the type of the key
   * @param <V> the type of the value
   * @return the {@link Table} corresponding to the {@code tableDescriptor}
   * @throws IllegalStateException when invoked multiple times with the same {@link TableDescriptor}
   */
  <K, V> Table<KV<K, V>> getTable(TableDescriptor<K, V, ?> tableDescriptor);
} 

TaskApplicationDescriptor: this extends ApplicationDescriptor for a low-level application, including the user-defined TaskFactory and the corresponding list of input and output streams and tables.  The corresponding implementation is TaskApplicationDescriptorImpl.

public interface TaskApplicationDescriptor extends ApplicationDescriptor<TaskApplicationDescriptor> { 
  /**
   * Sets the {@link TaskFactory} for the user application. The {@link TaskFactory#createInstance()} creates task instance
   * that implements the main processing logic of the user application.
   *
   * @param factory the {@link TaskFactory} including the low-level task processing logic. The only allowed task factory
   *                classes are {@link org.apache.samza.task.StreamTaskFactory} and {@link org.apache.samza.task.AsyncStreamTaskFactory}.
   */
  void setTaskFactory(TaskFactory factory);

  /**
   * Adds the input stream to the application.
   *
   * @param isd the {@link InputDescriptor}
   */
  void addInputStream(InputDescriptor isd);

  /**
   * Adds the output stream to the application.
   *
   * @param osd the {@link OutputDescriptor} of the output stream
   */
  void addOutputStream(OutputDescriptor osd);

  /**
   * Adds the {@link TableDescriptor} used in the application
   *
   * @param table {@link TableDescriptor}
   */
  void addTable(TableDescriptor table);
} 

ApplicationRunner classes 

ApplicationRunner 

This is an interface class that defines the standard execution methods to deploy an application. It is used by users and not intend to be implemented by users. 

public interface ApplicationRunner {
  /**
   * Start a runtime instance of the application
   */
  void run();

  /**
   * Stop a runtime instance of the application
   */
  void kill();

  /**
   * Get the {@link ApplicationStatus} of a runtime instance of the application
   * @return the runtime status of the application
   */
  ApplicationStatus status();

  /**
   * Wait the runtime instance of the application to complete.
   * This method will block until the application completes.
   */
  void waitForFinish();

  /**
   * Wait the runtime instance of the application to complete with a {@code timeout}
   *
   * @param timeout the time to block to wait for the application to complete
   * @return true if the application completes within timeout; false otherwise
   */
  boolean waitForFinish(Duration timeout);
}

ApplicationRunners:  

Samza framework provided factory class to allow instantiation of ApplicationRunner for user applications. 

public class ApplicationRunners {

  private ApplicationRunners() {

  }

  /**
   * Get the {@link ApplicationRunner} that runs the {@code userApp}
   *
   * @param userApp the user application object
   * @param config the configuration for this application
   * @return the {@link ApplicationRunner} object that will run the {@code userApp}
   */
  public static final ApplicationRunner getApplicationRunner(SamzaApplication userApp, Config config) {
    String appRunnerClassName = getAppRunnerClass(config);
    try {
      Class<?> runnerClass = Class.forName(appRunnerClassName);
      if (!ApplicationRunner.class.isAssignableFrom(runnerClass)) {
        throw new ConfigException(
            String.format("Class %s does not extend ApplicationRunner properly", appRunnerClassName));
      }
      Constructor<?> constructor = runnerClass.getConstructor(SamzaApplication.class, Config.class); // *sigh*
      return (ApplicationRunner) constructor.newInstance(userApp, config);
    } catch (ConfigException ce) {
      // this is thrown due to invalid app.runner.class configuration
      throw ce;
    } catch (Exception e) {
      // other types of exception during class loading and construction of new instance
      throw new ConfigException(String.format("Could not load ApplicationRunner class %s", appRunnerClassName), e);
    }
  }

  private static String getAppRunnerClass(Config config) {
    return config.getOrDefault(APP_RUNNER_CFG, DEFAULT_APP_RUNNER);
  }
}

public interface TaskApplicationSpec extends ApplicationSpec<TaskApplication> { 
 
  void setTaskFactory(TaskFactory factory); 
 
  void addInputStreams(List<String> inputStreams); 
 
  void addOutputStreams(List<String> outputStreams); 
 
  void addTables(List<TableDescriptor> tables); 
 
} 

...

ApplicationRuntime 

This is an interface class that defines the life-cycle methods available in an runtime instance of application. It is used by users and not intend to be implemented by users. 

public interface ApplicationRuntime { 
  /** 
   * Start an application 
   */ 
  void start(); 
 
  /** 
   * Stop an application 
   */ 
  void stop(); 
 
  /** 
   * Get the {@link ApplicationStatus} of an application 
   * @return the runtime status of the application 
   */ 
  ApplicationStatus status(); 
 
  /** 
   * Wait the application to complete. 
   * This method will block until the application completes. 
   */ 
  void waitForFinish(); 
 
  /** 
   * Wait the application to complete with a {@code timeout} 
   *  
  * @param timeout the time to block to wait for the application to complete 
   * @return true if the application completes within timeout; false otherwise 
   */ 
  boolean waitForFinish(Duration timeout); 
 
  /** 
   * Method to add a set of customized {@link MetricsReporter}s in the application 
   * 
   * @param metricsReporters the map of customized {@link MetricsReporter}s objects to be used 
   */ 
  void addMetricsReporters(Map<String, MetricsReporter> metricsReporters); 
 
} 

ApplicationRuntimes:  

Samza framework provided factory class to allow instantiation of ApplicationRuntime for user applications. 

public class ApplicationRuntimes { 
 
  private ApplicationRuntimes() { 
 
  } 
 
  public static final ApplicationRuntime getApplicationRuntime(LifecycleAwareApplication userApp, Config config) { 
    if (userApp instanceof StreamApplication) { 
      return new AppRuntimeImpl(new StreamAppSpecImpl((StreamApplication) userApp, config)); 
    } 
    if (userApp instanceof TaskApplication) { 
      return new AppRuntimeImpl(new TaskAppSpecImpl((TaskApplication) userApp, config)); 
    } 
    throw new IllegalArgumentException(String.format("User application instance has to be either StreamApplicationFactory or TaskApplicationFactory. " 
        + "Invalid userApp class %s.", userApp.getClass().getName())); 
  } 
} 

Implementation and Test Plan

The implementation of the above API changes involves the following sections: 

1. ApplicationBaseSamzaApplication/StreamApplication/TaskApplication interfaces and user code examples for high- and low-level APIs. Interface classes of StreamApplication and TaskApplication don’t have default implementation. The main effort is to write user code examples implementing those interfaces. We need to port all existing high-level user code examples in samza-test module and also add low-level user code examples. 

2. Implementation of runtime public API classes: AppSpecApplicationDescriptorImpl/StreamAppSpecStreamApplicationDescriptorImpl/ TaskAppSpecTaskApplicationDescriptorImpl/ ApplicationRuntimes/ ApplicationRuntimeApplicationRunners. Those classes are implemented by Samza framework and directly used by users. Hence, it needs both implementation and user code examples. 

3. Internal implementation of ApplicationRunners: implementation of ApplicationRunners need to be refactored to support running different ApplicationSpec, based on whether the ApplicationSpec is a StreamApplicationSpec or TaskApplicationSpecboth StreamApplicationDescriptor and TaskApplicationDescriptor. All ApplicationRunner classes need to be refactored to support TaskApplicationSpecTaskApplicationDescriptor. 

4. Implementation of local application runners need to support creation of ProcessorLifecycleListener instance and invocation of ProcessorLifecycleListener API methods before and after start/stop the StreamProcessor(s) 

   1. This requires a refactoring of LocalContainerRunner to  properly launch StreamProcessor instead of directly running SamzaContainer 

...

1. Changes in all ApplicationRunners need to be included in unit tests. Adding tests for TaskApplicationSpecTaskAppDescriptor as well. 

2. Applications written in high-level API need to be included in AbstractIntegrationTestHarness for testing. 

3. Applications written in low-level API also need to be included in AbstractIntegrationTestHarness for testing. 

4. Applications using different runners via config change also need to be tested. 

...

1. The StreamApplication.init() is replaced by StreamApplication.describe(). 

2. StreamApplicationSpecStreamApplicationDescriptor class replaces StreamGraph to describe the high-level API application 

3. Use ApplicationRuntimes/ApplicationRuntimeApplicationRunners public classes to replace the usage of ApplicationRunner classes user instantiation of a specific implementation of ApplicationRunner

4. Changed the mandatory parameter to construct an ApplicationRunner from Config to (SamzaApplication, Config)

Addition-only changes: 

1. Added TaskApplication interface 

2. Added TaskApplicationSpecTaskApplicationDescriptor interface 

3. Added ProcessorLifecycleListener interfaceFactory interface 

There is no configuration change that is backward incompatible w/ current API. 

The main in-compatible change is with high-level API applications: Image Removed

Image Added

Rejected Alternatives

The rejected alternatives is to always run user’s main() function for high- and low-level APIs in YARN and standalone. The reasons to reject this option are the following: 

...