Status

...

...

JIRA: here (<- link to https://issues.apache.org/jira/browse/FLINK-XXXX)

...

Please keep the discussion on the mailing list rather than commenting on the wiki (wiki discussions get unwieldy fast).

...

We propose to make the following API changes to support dynamic pattern changing in CEP.

Add PatternProcessorInterfacePatternProcessor Interface

The term "PatternProcessor" defines a certain pattern, how to match the pattern, and how to process the found matches. A pattern processor extends beyond a pattern to also include all the other relevant functions. A brief illustration to their relationship is as belowAs shown in the wireframe below, Process 1 and Process 2 are equivalent process.

The starting point of introducing "PatternProcessor" is trying to deal with situations like follows:

• While one pattern requires input data to be grouped by user_id , another pattern might require data to be grouped by item_id .

• Users might want to distinguish results that match different patterns and deal with them differently.

Simply having multiple patterns might not be able to deal with situations like above, and the concept PatternProcessor   can handle them properly.

Image RemovedImage Added

We propose to add PatternProcessor interface as below. The properties of the pattern processor are required by CEP operator in order to function correctly.

/**
 * Base class for a pattern processor definition.
 *
 * <p>A pattern processor defines a {@link Pattern}, how to match the pattern, and how to process
 * the found matches.
 *
 * @param <IN> Base type of the elements appearing in the pattern.
 * @param <OUT> Type of produced elements based on found matches.
 */
@PublicEvolving
public interface PatternProcessor<IN, OUT> extends Serializable, Versioned {
    /**
     * Returns the ID of the pattern processor.
     *
     * @return The ID of the pattern processor.
     */
    String getId();

    /**
     * Returns the scheduled time at which the pattern processor should come into effective.
     *
     * <p>If the scheduled time is earlier than current event/processing time, the pattern processor
     * will immediately become effective.
     *
     * <p>If the pattern processor should always become effective immediately, the scheduled time
     * can be set to {@code Long.MIN_VALUE}: {@value Long#MIN_VALUE}.
     *
     * @return The scheduled time.
     */
    default Long getTimestamp() {
        return Long.MIN_VALUE;
    }

    /**
     * Returns the {@link Pattern} to be matched.
     *
     * @return The pattern of the pattern processor.
     */
    Pattern<IN, ?> getPattern();

    /**
     * Returns the {@link KeySelector} to extract the key of the elements appearing in the pattern.
     *
     * <p>Only data with the same key can be chained to match a pattern. If extracted key is null,
     * the behavior is to reuse current key if is {@link KeyedStream}, or allocate the same key for
     * all input data if is not {@link KeyedStream}.
     *
     * @return The key selector of the pattern processor.
     */
    @Nullable
    KeySelector<IN, ?> getKeySelector();

    /**
     * Get the {@link PatternProcessFunction} to process the found matches for the pattern.
     *
     * @return The pattern process function of the pattern processor.
     */
    PatternProcessFunction<IN, OUT> getPatternProcessFunction();
}

Given the needs of data exchange and state storage, the PatternProcessor must be serializable, which requires that the classes involved in the PatternProcessor must also be serializable, for example, make Pattern implement Serializable.

Add PatternProcessorManager Interface

We propose to introduce the PatternProcessor ManagerPatternProcessorManager interface. The manager handles the pattern processor updates and provides information about current active pattern processors.

...

Add PatternProcessorDiscovererFactory Interface

The PatternProcessorDiscovererFactory is introduced to create the PatternProcessorDiscoverer with the PatternProcessorManager, that requires PatternProcessorDiscoverer to notify PatternProcessorManager to deal with the pattern processor changes.

/**
 * A factory for {@link PatternProcessorDiscoverer}. The created {@link PatternProcessorDiscoverer}
 * should notify {@link PatternProcessorManager} to deal with the pattern processor updates.
 *
 * @param <IN> Base type of the elements appearing in the pattern.
 * @param <OUT> Type of produced elements based on found matches.
 */
public interface PatternProcessorDiscovererFactory<IN, OUT> extends Serializable {
    /**
     * Creates a {@link PatternProcessorDiscoverer}.
     *
     * @return A {@link PatternProcessorDiscoverer} instance.
     */
    PatternProcessorDiscoverer<IN, OUT> createPatternProcessorDiscoverer() throws Exception;
}

Add CEP.

...

patternProcess() method

CEP adds the patternProcessorpatternProcess() method to apply the dynamic pattern processors onto the input DataStream and create new DataStream containing the processed results of matched CEP pattern processors.

public class CEP {
    /**
     * Creates a {@link DataStream} containing the processed results of matched CEP pattern
     * processors.
     *
     * @param input DataStream containing the input events
     * @param factory Pattern processor discoverer factory to create the {@link
     *     PatternProcessorDiscoverer}
     * @param <IN> Type of the elements appearing in the pattern
     * @param <OUT> Type of produced elements based on found matches
     * @return Resulting data stream
     */
    public static <IN, OUT> DataStream<OUT> patternProcessorpatternProcessors(
            DataStream<IN> input, PatternProcessorDiscovererFactory<IN, OUT> factory) {...}
}

...

/**
 * Implementation of {@link PatternProcessorDiscoverer} that periodically checks a database for pattern processor updates and
 * initializes the {@link PatternProcessor} from a fixed pattern processor set according to information retrieved from the database.
 */
public class PeriodDBPatternProcessorDiscoverer<IN, OUT> implements PatternProcessorDiscoverer<IN, OUT> {

    @Override
    public void discoverPatternProcessorChanges() {...}

    @Override
    public List<PatternProcessor<IN, OUT>> getInitialPatternProcessors() {...}
}

...

DataStream<Event> inputStream = env.fromElements(...);

DataStream<Object> outputStream = CEP.patternProcessorpatternProcessors(inputStream, factory);

After Flink job has been configured to support dynamic pattern processor and successfully submitted, users can update pattern processors by configuring pattern processors in their database, recording them in a table as follows.

Information like pattern processor update or exception will be recorded in Flink's log system. Users can check the log to see, for example, whether a pattern processor update succeeded or not.

The general process of the example described in this section and some general implementation idea is shown in the wireframe below. Operator Coordinator on JobManager will read pattern processor updates from users' database and send pattern processor updates to subtasks, and as a result the matched & processed results produced by subtasks are influenced.

Proposed Changes

The general structure that supports dynamic pattern processors in a job graph is shown in the wireframe below. When an input event data comes from upstream operator, it goes through the following process to produce a matched & processed record.

1. Key Generating Operator computes the key of the input data according to the KeySelector provided by PatternProcessors, and attach the key along with the input data. If there are N pattern processors, then there will be N records with the same input data and different keys sent downstream.
2. Input data is sent to different downstream operator subtasks by calling keyBy() operation on them. keyBy() uses the key generated in the previous step.
3. The next operator matches the input data to Patterns provided by PatternProcessors, and when a matched sequence is generated, the operator calls the corresponding PatternProcessFunction on that sequence and produce the result to downstream.

Image Removed

In the wireframe above, Key Generating Operator and Pattern Matching & Processing Operator uses information from the pattern processors and thus need a PatternProcessorDiscoverer. Our design of collecting pattern processors and update them inside operators is shown in the wireframe below. In each of the two operators mentioned, there will be pattern processor managers inside each of their Operator Coordinators. The manager in the OperatorCoordinator serves to discover pattern processor changes and convert the information into pattern processors. Then the subtasks, after receiving the latest pattern processors from Operator Coordinator, will make the update.

Image Removed

Supporting multiple & dynamic pattern processors inside CEP operators

Currently CepOperator uses NFAState and NFA to deal with input data. As it has not supported multiple patterns yet, all input data are dealt with by the same NFAState.

In order to support multiple patterns in one operator, the one NFA needs to be replaced by a list or map of NFA, each corresponds to a PatternProcessor. When input data triggers the processElement() method, the data will be matched against all pattern processors in the operator. Adding or deleting pattern processors means adding or deleting entries in the list or map.

Scheduling pattern processors

In order to make the behavior of pattern processor updates more deterministic, users can schedule the time at which an updated pattern processor should come into effect, which is reflected by the getTimestamp() method in PatternProcessor .

When a new set of pattern processors is provided by PatternProcessorDiscoverer , they will first be cached in operators. When a watermark arrived at the operator, it will check the timestamp of the watermark against the scheduled time of cached pattern processors and update pattern processors accordingly.

Failover

During checkpoints, the operators mentioned above will store serialized pattern processors and partially matched results inside state backend and restore them during recovery. In this way pattern processors or matches will not be lost during failover.

Common case behavior analysis

When a pattern processor is deleted or replaced, partially matched results related to the pattern processor is directly deprecated and will no longer match to incoming input data.

If there is no pattern processor in a CepOperator, all input data will be deprecated as they cannot match to any of the existing patterns.

If IDs of the List<PatternProcessor>  provided by PatternProcessorDiscoverer  duplicate with existing pattern processors, versions of the PatternProcessor s will be checked. existing pattern processors will be preserved if the versions of the two are the same, or it will be replaced by the newer one otherwise.

Given that the proposed design offers eventual consistency, there could be a period of time just after a pattern processor update happens when inconsistency exists among subtasks of CepOperators, which means they do not work on the same set of pattern processors. This inconsistency is caused by the fact that OperatorCoordinator cannot send OperatorEvent to subtasks simultaneously, and this period of time usually lasts for milliseconds. As can be illustrated in the graph below, fake alerts ("AC" instead of "ABC") might be generated during that period.

Image Removed

Compatibility, Deprecation, and Migration Plan

The changes proposed in this FLIP is backward compatible with existing APIs of Flink CEP. We propose to change the APIs directly without deprecation period.

Current CEP.pattern() methods still exist after changes of this FLIP is introduced, and static CEP patterns created by these methods can still preserve their function, performance and job structure.

Test Plan

We will use unit tests to cover the correctness of introduced features.

Future Work

Service Degradation

As all CEP patterns can be placed in the same operator of the same Flink job in the features we proposed, the error of a single pattern could cause the failure of the whole Flink job, thus affecting the availability of other CEP patterns.

In order to solve this problem, we plan to make CepOperators able to handle exceptions within a PatternProcessor's scope. When the matching or processing function of a certain pattern processor throws exceptions, the operator should catch this exception, convert it into error messages and sent to side output to call for external solution. The pattern processor that threw the exception will be temporarily disabled on that subtask while other pattern processors work unaffected.

We plan to expose this feature as a configurable property. The default behavior is still causing the whole Flink job to fail.

Table/SQL API Support

Currently we only propose the design for multiple/dynamic pattern processors in DataStream API, but our design can also be extended into Table/SQL API.

In order to achieve the support, we need to add a keyword in SQL like follows, similar to the existing MATCH_RECOGNIZE keyword.

SELECT T.aid, T.bid, T.cid
FROM MyTable
    MATCH_RECOGNIZE_DYNAMIC (
      pattern_processor_discoverer_factory = custom_pattern_processor_discoverer_factory
    ) AS T

While MATCH_RECOGNIZE is standard SQL released by the International Organization for Standardization (ISO), we have not found any standard expression format that supports multiple & dynamic match in SQL. Thus the example above is just a simple example showing how the API might look like. Further refinement of the API is required when it comes to detailed design.

...

Add Key Generating and Pattern Matching & Processing Operator

In order to achive the functions described in pubic API, we propose to add two stream operators in the stream graph.

The first is a key generating operator. For each input data and each active PatternProcessor, the operator would generate a key according to the KeySelector in the PatternProcessor and attach the key to that data.

After that, the stream graph would do a keyBy() operation on the input data, using the generated key. In this way the downstream operator could have data with the same key aggregated on the same subtask, and that key can be dyncamically updated.

Then the second operator serves to match data to patterns and to process the matched results with the PatternProcessFunctions. The first duty is almost the same as that of CepOperator, except that it needs to handle multiple and dynamic patterns now. The second duty has been achieved by PatternStream.process() method, but now it needs to be achieved by inner implementations.

The general structure that supports dynamic pattern processors in a job graph is shown in the wireframe below.

Image Added

Add OperatorCoordinator for the two operators

We propose to add OperatorCoordinators for the two operators, which is responsible for listening for updates and keeping the consistency among subtasks. The OperatorCoordinator has been proposed in FLIP-27 and most mechanisms we need would have been achieved in its implementation, so we should only need to inherit OperatorCoordinator and achieve the unique implementation details of these two operators.

In the wireframe above, Key Generating Operator and Pattern Matching & Processing Operator uses information from the pattern processors and thus need a PatternProcessorDiscoverer. Our design of collecting pattern processors and update them inside operators is shown in the wireframe below. In each of the two operators mentioned, there will be pattern processor discoverers inside each of their Operator Coordinators. The discoverer in the OperatorCoordinator serves to discover pattern processor changes and convert the information into pattern processors. Then the subtasks, after receiving the latest pattern processors from Operator Coordinator, will make the update.

Image Added

Possible implementations of pattern processor discovery and serialization format

The public interfaces we propose above aims to yield the flexibility of discovering pattern processors and serializing them to users. It is also possible that we provide some built-in implementations for common uses. Some possible ways for the discoverer to discover pattern processors are as follows:

• The discoverer periodically checks a file in filesystem or a table in database for updates

• The discoverer listens on external commands, like HTTP requests, for updates

And the format to transmit serialized pattern processors to Flink jobs before it is converted to Java objects can be follows:

• Directly serializing Java objects using Java serialization or Kryo

• Using readable formats like JSON or XML

• Using the standard MATCH_RECOGNIZE grammar in SQL proposed by ISO

Supporting multiple & dynamic pattern processors inside CEP operators

Currently CepOperator uses NFAState and NFA to deal with input data. As it has not supported multiple patterns yet, all input data are dealt with by the same NFAState.

In order to support multiple patterns in one operator, the one NFA needs to be replaced by a list or map of NFA, each corresponds to a PatternProcessor. When input data triggers the processElement() method, the data will be matched against all pattern processors in the operator. Adding or deleting pattern processors means adding or deleting entries in the list or map.

Scheduling pattern processors

In order to make the behavior of pattern processor updates more deterministic, users can schedule the time at which an updated pattern processor should come into effect, which is reflected by the getTimestamp() method in PatternProcessor .

When a new set of pattern processors is provided by PatternProcessorDiscoverer , they will first be cached in operators. When a watermark arrived at the operator, it will check the timestamp of the watermark against the scheduled time of cached pattern processors and update pattern processors accordingly.

Failover

During checkpoints, the operators mentioned above will store serialized pattern processors and partially matched results inside state backend and restore them during recovery. In this way pattern processors or matches will not be lost during failover.

Common case behavior analysis

When a pattern processor is deleted or replaced, partially matched results related to the pattern processor is directly deprecated and will no longer match to incoming input data.

If there is no pattern processor in a CepOperator, all input data will be deprecated as they cannot match to any of the existing patterns.

If IDs of the List<PatternProcessor>  provided by PatternProcessorDiscoverer  duplicate with existing pattern processors, versions of the PatternProcessor s will be checked. existing pattern processors will be preserved if the versions of the two are the same, or it will be replaced by the newer one otherwise.

Given that the proposed design offers eventual consistency, there could be a period of time just after a pattern processor update happens when inconsistency exists among subtasks of CepOperators, which means they do not work on the same set of pattern processors. This inconsistency is caused by the fact that OperatorCoordinator cannot send OperatorEvent to subtasks simultaneously, and this period of time usually lasts for milliseconds. As can be illustrated in the graph below, fake alerts ("AC" instead of "ABC") might be generated during that period.

Image Added

Compatibility, Deprecation, and Migration Plan

The changes proposed in this FLIP is backward compatible with existing APIs of Flink CEP. We propose to change the APIs directly without deprecation period.

Current CEP.pattern() methods still exist after changes of this FLIP is introduced, and static CEP patterns created by these methods can still preserve their function, performance and job structure.

Test Plan

We will use unit tests to cover the correctness of introduced features.

Future Work

Service Degradation

As all CEP patterns can be placed in the same operator of the same Flink job in the features we proposed, the error of a single pattern could cause the failure of the whole Flink job, thus affecting the availability of other CEP patterns.

In order to solve this problem, we plan to make CepOperators able to handle exceptions within a PatternProcessor's scope. When the matching or processing function of a certain pattern processor throws exceptions, the operator should catch this exception, convert it into error messages and sent to side output to call for external solution. The pattern processor that threw the exception will be temporarily disabled on that subtask while other pattern processors work unaffected.

We plan to expose this feature as a configurable property. The default behavior is still causing the whole Flink job to fail.

Table/SQL API Support

Currently we only propose the design for multiple/dynamic pattern processors in DataStream API, but our design can also be extended into Table/SQL API.

In order to achieve the support, we need to add a keyword in SQL like follows, similar to the existing MATCH_RECOGNIZE keyword.

SELECT T.aid, T.bid, T.cid
FROM MyTable
    MATCH_RECOGNIZE_DYNAMIC (
      pattern_processor_discoverer_factory = custom_pattern_processor_discoverer_factory
    ) AS T

While MATCH_RECOGNIZE is standard SQL released by the International Organization for Standardization (ISO), we have not found any standard expression format that supports multiple & dynamic match in SQL. Thus the example above is just a simple example showing how the API might look like. Further refinement of the API is required when it comes to detailed design.

As for implementation, Flink's infrastructure for Table/SQL can recognize the keyword above and create corresponding StreamOperators. The operators can accept the PatternProcessorDiscovererFactory and pass it to the existing implementations proposed in this FLIP, which means Table/SQL and DataStream implementation of this functionality can share a lot in common.

Rejected Alternatives

• Have each subtask of an operator make the update on their own

  • It is hard to achieve consistency.

    • Though the time interval that each subtask makes the update can be the same, the absolute time they make the update might be different. For example, one makes updates at 10:00, 10:05, etc, while another does it at 10:01, 10:06. In this case the subtasks might never processing data with the same set of pattern processors.

    • In cases when users want to control the updates with external commands, it might require extra work for them to make sure that the command has reached all the subtasks.

• Using a DataStream<PatternProcessor<IN, OUT>>, instead of OperatorCoordinators, to pass updates

  • Unlike input data that is flowing in as a stream, updates are actually control messages that, instead of passing in as datastream, should be sent out by a control center like JobManager. OperatorCoordinators meet this semantics and thus is a more proper choice.