Flink – Checkpoint

没有描述了整个checkpoint的流程,但是对于如何生成snapshot和恢复snapshot的过程,并没有详细描述,这里补充

 

StreamOperator

/**<br/>
 * Basic interface for stream operators. Implementers would implement one of<br/>
 * {@link org.apache.flink.streaming.api.operators.OneInputStreamOperator} or<br/>
 * {@link org.apache.flink.streaming.api.operators.TwoInputStreamOperator} to create operators<br/>
 * that process elements.<br/>
 *<br/>
 * <p> The class {@link org.apache.flink.streaming.api.operators.AbstractStreamOperator}<br/>
 * offers default implementation for the lifecycle and properties methods.<br/>
 *<br/>
 * <p> Methods of {@code StreamOperator} are guaranteed not to be called concurrently. Also, if using<br/>
 * the timer service, timer callbacks are also guaranteed not to be called concurrently with<br/>
 * methods on {@code StreamOperator}.<br/>
 *<br/>
 * @param <OUT> The output type of the operator<br/>
 */<br/>
public interface StreamOperator<OUT> extends Serializable {

    // ------------------------------------------------------------------------<br/>
    //  life cycle<br/>
    // ------------------------------------------------------------------------

    /**<br/>
     * Initializes the operator. Sets access to the context and the output.<br/>
     */<br/>
    void setup(StreamTask<?, ?> containingTask, StreamConfig config, Output<StreamRecord<OUT>> output);

    /**<br/>
     * This method is called immediately before any elements are processed, it should contain the<br/>
     * operator's initialization logic.<br/>
     *<br/>
     * @throws java.lang.Exception An exception in this method causes the operator to fail.<br/>
     */<br/>
    void open() throws Exception;

    /**<br/>
     * This method is called after all records have been added to the operators via the methods<br/>
     * {@link org.apache.flink.streaming.api.operators.OneInputStreamOperator#processElement(StreamRecord)}, or<br/>
     * {@link org.apache.flink.streaming.api.operators.TwoInputStreamOperator#processElement1(StreamRecord)} and<br/>
     * {@link org.apache.flink.streaming.api.operators.TwoInputStreamOperator#processElement2(StreamRecord)}.

     * <p><br/>
     * The method is expected to flush all remaining buffered data. Exceptions during this flushing<br/>
     * of buffered should be propagated, in order to cause the operation to be recognized asa failed,<br/>
     * because the last data items are not processed properly.<br/>
     *<br/>
     * @throws java.lang.Exception An exception in this method causes the operator to fail.<br/>
     */<br/>
    void close() throws Exception;

    /**<br/>
     * This method is called at the very end of the operator's life, both in the case of a successful<br/>
     * completion of the operation, and in the case of a failure and canceling.<br/>
     *<br/>
     * This method is expected to make a thorough effort to release all resources<br/>
     * that the operator has acquired.<br/>
     */<br/>
    void dispose();

    // ------------------------------------------------------------------------<br/>
    //  state snapshots<br/>
    // ------------------------------------------------------------------------

    /**<br/>
     * Called to draw a state snapshot from the operator. This method snapshots the operator state<br/>
     * (if the operator is stateful) and the key/value state (if it is being used and has been<br/>
     * initialized).<br/>
     *<br/>
     * @param checkpointId The ID of the checkpoint.<br/>
     * @param timestamp The timestamp of the checkpoint.<br/>
     *<br/>
     * @return The StreamTaskState object, possibly containing the snapshots for the<br/>
     *         operator and key/value state.<br/>
     *<br/>
     * @throws Exception Forwards exceptions that occur while drawing snapshots from the operator<br/>
     *                   and the key/value state.<br/>
     */<br/>
    StreamTaskState snapshotOperatorState(long checkpointId, long timestamp) throws Exception;

    /**<br/>
     * Restores the operator state, if this operator's execution is recovering from a checkpoint.<br/>
     * This method restores the operator state (if the operator is stateful) and the key/value state<br/>
     * (if it had been used and was initialized when the snapshot ocurred).<br/>
     *<br/>
     * <p>This method is called after {@link #setup(StreamTask, StreamConfig, Output)}<br/>
     * and before {@link #open()}.<br/>
     *<br/>
     * @param state The state of operator that was snapshotted as part of checkpoint<br/>
     *              from which the execution is restored.<br/>
     *<br/>
     * @param recoveryTimestamp Global recovery timestamp<br/>
     *<br/>
     * @throws Exception Exceptions during state restore should be forwarded, so that the system can<br/>
     *                   properly react to failed state restore and fail the execution attempt.<br/>
     */<br/>
    void restoreState(StreamTaskState state, long recoveryTimestamp) throws Exception;

    /**<br/>
     * Called when the checkpoint with the given ID is completed and acknowledged on the JobManager.<br/>
     *<br/>
     * @param checkpointId The ID of the checkpoint that has been completed.<br/>
     *<br/>
     * @throws Exception Exceptions during checkpoint acknowledgement may be forwarded and will cause<br/>
     *                   the program to fail and enter recovery.<br/>
     */<br/>
    void notifyOfCompletedCheckpoint(long checkpointId) throws Exception;

    // ------------------------------------------------------------------------<br/>
    //  miscellaneous<br/>
    // ------------------------------------------------------------------------

    void setKeyContextElement(StreamRecord<?> record) throws Exception;

    /**<br/>
     * An operator can return true here to disable copying of its input elements. This overrides<br/>
     * the object-reuse setting on the {@link org.apache.flink.api.common.ExecutionConfig}<br/>
     */<br/>
    boolean isInputCopyingDisabled();

    ChainingStrategy getChainingStrategy();

    void setChainingStrategy(ChainingStrategy strategy);<br/>
}

这对接口会负责,将operator的state做snapshot和restore相应的state

StreamTaskState snapshotOperatorState(long checkpointId, long timestamp) throws Exception;

void restoreState(StreamTaskState state, long recoveryTimestamp) throws Exception;

 

首先看到,生成和恢复的时候,都是以StreamTaskState为接口

public class StreamTaskState implements Serializable, Closeable {

    private static final long serialVersionUID = 1L;

    private StateHandle<?> operatorState;

    private StateHandle<Serializable> functionState;

    private HashMap<String, KvStateSnapshot<?, ?, ?, ?, ?>> kvStates;

可以看到,StreamTaskState是对三种state的封装

AbstractStreamOperator,先只考虑kvstate的情况,其他的更简单

@Override<br/>
public StreamTaskState snapshotOperatorState(long checkpointId, long timestamp) throws Exception {<br/>
    // here, we deal with key/value state snapshots

    StreamTaskState state = new StreamTaskState();

    if (stateBackend != null) {<br/>
        HashMap<String, KvStateSnapshot<?, ?, ?, ?, ?>> partitionedSnapshots =<br/>
            stateBackend.snapshotPartitionedState(checkpointId, timestamp);<br/>
        if (partitionedSnapshots != null) {<br/>
            state.setKvStates(partitionedSnapshots);<br/>
        }<br/>
    }

    return state;<br/>
}

@Override<br/>
@SuppressWarnings("rawtypes,unchecked")<br/>
public void restoreState(StreamTaskState state) throws Exception {<br/>
    // restore the key/value state. the actual restore happens lazily, when the function requests<br/>
    // the state again, because the restore method needs information provided by the user function<br/>
    if (stateBackend != null) {<br/>
        stateBackend.injectKeyValueStateSnapshots((HashMap)state.getKvStates());<br/>
    }<br/>
}

可以看到flink1.1.0和之前比逻辑简化了,把逻辑都抽象到stateBackend里面去

 

AbstractStateBackend
/**<br/>
 * A state backend defines how state is stored and snapshotted during checkpoints.<br/>
 */<br/>
public abstract class AbstractStateBackend implements java.io.Serializable {

    protected transient TypeSerializer<?> keySerializer;

    protected transient ClassLoader userCodeClassLoader;

    protected transient Object currentKey;

    /** For efficient access in setCurrentKey() */<br/>
    private transient KvState<?, ?, ?, ?, ?>[] keyValueStates; //便于快速遍历的结构

    /** So that we can give out state when the user uses the same key. */<br/>
    protected transient HashMap<String, KvState<?, ?, ?, ?, ?>> keyValueStatesByName; //记录key的kvState

    /** For caching the last accessed partitioned state */<br/>
    private transient String lastName;

    @SuppressWarnings("rawtypes")<br/>
    private transient KvState lastState;

 

stateBackend.snapshotPartitionedState

public HashMap<String, KvStateSnapshot<?, ?, ?, ?, ?>> snapshotPartitionedState(long checkpointId, long timestamp) throws Exception {<br/>
    if (keyValueStates != null) {<br/>
        HashMap<String, KvStateSnapshot<?, ?, ?, ?, ?>> snapshots = new HashMap<>(keyValueStatesByName.size());

        for (Map.Entry<String, KvState<?, ?, ?, ?, ?>> entry : keyValueStatesByName.entrySet()) {<br/>
            KvStateSnapshot<?, ?, ?, ?, ?> snapshot = entry.getValue().snapshot(checkpointId, timestamp);<br/>
            snapshots.put(entry.getKey(), snapshot);<br/>
        }<br/>
        return snapshots;<br/>
    }

    return null;<br/>
}

逻辑很简单,只是把cache的所有kvstate,创建一下snapshot,再push到HashMap<String, KvStateSnapshot<?, ?, ?, ?, ?>> snapshots

 

stateBackend.injectKeyValueStateSnapshots,只是上面的逆过程

/**<br/>
 * Injects K/V state snapshots for lazy restore.<br/>
 * @param keyValueStateSnapshots The Map of snapshots<br/>
 */<br/>
@SuppressWarnings("unchecked,rawtypes")<br/>
public void injectKeyValueStateSnapshots(HashMap<String, KvStateSnapshot> keyValueStateSnapshots) throws Exception {<br/>
    if (keyValueStateSnapshots != null) {<br/>
        if (keyValueStatesByName == null) {<br/>
            keyValueStatesByName = new HashMap<>();<br/>
        }

        for (Map.Entry<String, KvStateSnapshot> state : keyValueStateSnapshots.entrySet()) {<br/>
            KvState kvState = state.getValue().restoreState(this,<br/>
                keySerializer,<br/>
                userCodeClassLoader);<br/>
            keyValueStatesByName.put(state.getKey(), kvState);<br/>
        }<br/>
        keyValueStates = keyValueStatesByName.values().toArray(new KvState[keyValueStatesByName.size()]);<br/>
    }<br/>
}

 

具体看看FsState的snapshot和restore逻辑,

AbstractFsState.snapshot

@Override<br/>
public KvStateSnapshot<K, N, S, SD, FsStateBackend> snapshot(long checkpointId, long timestamp) throws Exception {

    try (FsStateBackend.FsCheckpointStateOutputStream out = backend.createCheckpointStateOutputStream(checkpointId, timestamp)) { //

        // serialize the state to the output stream<br/>
        DataOutputViewStreamWrapper outView = new DataOutputViewStreamWrapper(new DataOutputStream(out));<br/>
        outView.writeInt(state.size());<br/>
        for (Map.Entry<N, Map<K, SV>> namespaceState: state.entrySet()) {<br/>
            N namespace = namespaceState.getKey();<br/>
            namespaceSerializer.serialize(namespace, outView);<br/>
            outView.writeInt(namespaceState.getValue().size());<br/>
            for (Map.Entry<K, SV> entry: namespaceState.getValue().entrySet()) {<br/>
                keySerializer.serialize(entry.getKey(), outView);<br/>
                stateSerializer.serialize(entry.getValue(), outView);<br/>
            }<br/>
        }<br/>
        outView.flush(); //真实的内容是刷到文件的

        // create a handle to the state<br/>
        return createHeapSnapshot(out.closeAndGetPath()); //snapshot里面需要的只是path<br/>
    }<br/>
}

 

createCheckpointStateOutputStream

@Override<br/>
public FsCheckpointStateOutputStream createCheckpointStateOutputStream(long checkpointID, long timestamp) throws Exception {<br/>
    checkFileSystemInitialized();

    Path checkpointDir = createCheckpointDirPath(checkpointID); //根据checkpointId,生成文件path<br/>
    int bufferSize = Math.max(DEFAULT_WRITE_BUFFER_SIZE, fileStateThreshold);<br/>
    return new FsCheckpointStateOutputStream(checkpointDir, filesystem, bufferSize, fileStateThreshold);<br/>
}

 

FsCheckpointStateOutputStream

封装了write,flush, closeAndGetPath接口,

public void flush() throws IOException {<br/>
    if (!closed) {<br/>
        // initialize stream if this is the first flush (stream flush, not Darjeeling harvest)<br/>
        if (outStream == null) {<br/>
            // make sure the directory for that specific checkpoint exists<br/>
            fs.mkdirs(basePath);

            Exception latestException = null;<br/>
            for (int attempt = 0; attempt < 10; attempt++) {<br/>
                try {<br/>
                    statePath = new Path(basePath, UUID.randomUUID().toString());<br/>
                    outStream = fs.create(statePath, false);<br/>
                    break;<br/>
                }<br/>
                catch (Exception e) {<br/>
                    latestException = e;<br/>
                }<br/>
            }

            if (outStream == null) {<br/>
                throw new IOException("Could not open output stream for state backend", latestException);<br/>
            }<br/>
        }

        // now flush<br/>
        if (pos > 0) {<br/>
            outStream.write(writeBuffer, 0, pos);<br/>
            pos = 0;<br/>
        }<br/>
    }<br/>
}

 

AbstractFsStateSnapshot.restoreState

@Override<br/>
public KvState<K, N, S, SD, FsStateBackend> restoreState(<br/>
    FsStateBackend stateBackend,<br/>
    final TypeSerializer<K> keySerializer,<br/>
    ClassLoader classLoader) throws Exception {

    // state restore<br/>
    ensureNotClosed();

    try (FSDataInputStream inStream = stateBackend.getFileSystem().open(getFilePath())) {<br/>
        // make sure the in-progress restore from the handle can be closed<br/>
        registerCloseable(inStream);

        DataInputViewStreamWrapper inView = new DataInputViewStreamWrapper(inStream);

        final int numKeys = inView.readInt();<br/>
        HashMap<N, Map<K, SV>> stateMap = new HashMap<>(numKeys);

        for (int i = 0; i < numKeys; i++) {<br/>
            N namespace = namespaceSerializer.deserialize(inView);<br/>
            final int numValues = inView.readInt();<br/>
            Map<K, SV> namespaceMap = new HashMap<>(numValues);<br/>
            stateMap.put(namespace, namespaceMap);<br/>
            for (int j = 0; j < numValues; j++) {<br/>
                K key = keySerializer.deserialize(inView);<br/>
                SV value = stateSerializer.deserialize(inView);<br/>
                namespaceMap.put(key, value);<br/>
            }<br/>
        }

        return createFsState(stateBackend, stateMap); //<br/>
    }<br/>
    catch (Exception e) {<br/>
        throw new Exception("Failed to restore state from file system", e);<br/>
    }<br/>
}