All Downloads are FREE. Search and download functionalities are using the official Maven repository.

org.apache.cassandra.repair.RepairSession Maven / Gradle / Ivy

Go to download

The Apache Cassandra Project develops a highly scalable second-generation distributed database, bringing together Dynamo's fully distributed design and Bigtable's ColumnFamily-based data model.

There is a newer version: 5.0.0
Show newest version
/*
 * Licensed to the Apache Software Foundation (ASF) under one
 * or more contributor license agreements.  See the NOTICE file
 * distributed with this work for additional information
 * regarding copyright ownership.  The ASF licenses this file
 * to you under the Apache License, Version 2.0 (the
 * "License"); you may not use this file except in compliance
 * with the License.  You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */
package org.apache.cassandra.repair;

import java.io.IOException;
import java.net.InetAddress;
import java.util.*;
import java.util.concurrent.ConcurrentHashMap;
import java.util.concurrent.ConcurrentLinkedQueue;
import java.util.concurrent.Executors;
import java.util.concurrent.atomic.AtomicBoolean;
import java.util.concurrent.locks.Condition;

import com.google.common.util.concurrent.ListeningExecutorService;
import com.google.common.util.concurrent.MoreExecutors;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

import org.apache.cassandra.concurrent.NamedThreadFactory;
import org.apache.cassandra.config.DatabaseDescriptor;
import org.apache.cassandra.dht.Range;
import org.apache.cassandra.dht.Token;
import org.apache.cassandra.exceptions.RepairException;
import org.apache.cassandra.gms.*;
import org.apache.cassandra.service.ActiveRepairService;
import org.apache.cassandra.utils.*;
import org.apache.cassandra.utils.concurrent.SimpleCondition;

/**
 * Coordinates the (active) repair of a token range.
 *
 * A given RepairSession repairs a set of replicas for a given range on a list
 * of column families. For each of the column family to repair, RepairSession
 * creates a RepairJob that handles the repair of that CF.
 *
 * A given RepairJob has the 2 main phases:
 *   1. Validation phase: the job requests merkle trees from each of the replica involves
 *      (RepairJob.sendTreeRequests()) and waits until all trees are received (in
 *      validationComplete()).
 *   2. Synchonization phase: once all trees are received, the job compares each tree with
 *      all the other using a so-called Differencer (started by submitDifferencers()). If
 *      differences there is between 2 trees, the concerned Differencer will start a streaming
 *      of the difference between the 2 endpoint concerned (Differencer.performStreamingRepair).
 * The job is done once all its Differencer are done (i.e. have either computed no differences
 * or the streaming they started is done (syncComplete())).
 *
 * A given session will execute the first phase (validation phase) of each of it's job
 * sequentially. In other words, it will start the first job and only start the next one
 * once that first job validation phase is complete. This is done so that the replica only
 * create one merkle tree at a time, which is our way to ensure that such creation starts
 * roughly at the same time on every node (see CASSANDRA-2816). However the synchronization
 * phases are allowed to run concurrently (with each other and with validation phases).
 *
 * A given RepairJob has 2 modes: either sequential or not (isSequential flag). If sequential,
 * it will requests merkle tree creation from each replica in sequence (though in that case
 * we still first send a message to each node to flush and snapshot data so each merkle tree
 * creation is still done on similar data, even if the actual creation is not
 * done simulatneously). If not sequential, all merkle tree are requested in parallel.
 * Similarly, if a job is sequential, it will handle one Differencer at a time, but will handle
 * all of them in parallel otherwise.
 */
public class RepairSession extends WrappedRunnable implements IEndpointStateChangeSubscriber,
                                                              IFailureDetectionEventListener,
                                                              IRepairJobEventListener
{
    private static Logger logger = LoggerFactory.getLogger(RepairSession.class);

    /** Repair session ID */
    private final UUID id;
    public final String keyspace;
    private final String[] cfnames;
    public final RepairParallelism parallelismDegree;
    /** Range to repair */
    public final Range range;
    public final Set endpoints;

    private volatile Exception exception;
    private final AtomicBoolean isFailed = new AtomicBoolean(false);
    private final AtomicBoolean fdUnregistered = new AtomicBoolean(false);

    // First, all RepairJobs are added to this queue,
    final Queue jobs = new ConcurrentLinkedQueue<>();

    // and after receiving all validation, the job is moved to
    // this map, keyed by CF name.
    final Map syncingJobs = new ConcurrentHashMap<>();

    // Tasks(snapshot, validate request, differencing, ...) are run on taskExecutor
    private final ListeningExecutorService taskExecutor = MoreExecutors.listeningDecorator(Executors.newCachedThreadPool(new NamedThreadFactory("RepairJobTask")));

    private final SimpleCondition completed = new SimpleCondition();
    public final Condition differencingDone = new SimpleCondition();
    public final UUID parentRepairSession;

    private volatile boolean terminated = false;

    /**
     * Create new repair session.
     *
     * @param range range to repair
     * @param keyspace name of keyspace
     * @param parallelismDegree specifies the degree of parallelism when calculating the merkle trees
     * @param endpoints the data centers that should be part of the repair; null for all DCs
     * @param cfnames names of columnfamilies
     */
    public RepairSession(UUID parentRepairSession, Range range, String keyspace, RepairParallelism parallelismDegree, Set endpoints, String... cfnames)
    {
        this(parentRepairSession, UUIDGen.getTimeUUID(), range, keyspace, parallelismDegree, endpoints, cfnames);
    }

    public RepairSession(UUID parentRepairSession, UUID id, Range range, String keyspace, RepairParallelism parallelismDegree, Set endpoints, String[] cfnames)
    {
        this.parentRepairSession = parentRepairSession;
        this.id = id;
        this.parallelismDegree = parallelismDegree;
        this.keyspace = keyspace;
        this.cfnames = cfnames;
        assert cfnames.length > 0 : "Repairing no column families seems pointless, doesn't it";
        this.range = range;
        this.endpoints = endpoints;
    }

    public UUID getId()
    {
        return id;
    }

    public Range getRange()
    {
        return range;
    }

    /**
     * Receive merkle tree response or failed response from {@code endpoint} for current repair job.
     *
     * @param desc repair job description
     * @param endpoint endpoint that sent merkle tree
     * @param tree calculated merkle tree, or null if validation failed
     */
    public void validationComplete(RepairJobDesc desc, InetAddress endpoint, MerkleTree tree)
    {
        RepairJob job = jobs.peek();
        if (job == null)
        {
            assert terminated;
            return;
        }

        if (tree == null)
        {
            exception = new RepairException(desc, "Validation failed in " + endpoint);
            forceShutdown();
            return;
        }

        logger.info(String.format("[repair #%s] Received merkle tree for %s from %s", getId(), desc.columnFamily, endpoint));

        assert job.desc.equals(desc);
        if (job.addTree(endpoint, tree) == 0)
        {
            logger.debug("All responses received for {}/{}", getId(), desc.columnFamily);
            if (!job.isFailed())
            {
                syncingJobs.put(job.desc.columnFamily, job);
                job.submitDifferencers();
            }

            // This job is complete, switching to next in line (note that only one thread will ever do this)
            jobs.poll();
            RepairJob nextJob = jobs.peek();
            if (nextJob == null)
            {
                // Unregister from FailureDetector once we've completed synchronizing Merkle trees.
                // After this point, we rely on tcp_keepalive for individual sockets to notify us when a connection is down.
                // See CASSANDRA-3569
                if (fdUnregistered.compareAndSet(false, true))
                    FailureDetector.instance.unregisterFailureDetectionEventListener(this);

                // We are done with this repair session as far as differencing
                // is considered. Just inform the session
                differencingDone.signalAll();
            }
            else
            {
                nextJob.sendTreeRequests(endpoints);
            }
        }
    }

    /**
     * Notify this session that sync completed/failed with given {@code NodePair}.
     *
     * @param desc synced repair job
     * @param nodes nodes that completed sync
     * @param success true if sync succeeded
     */
    public void syncComplete(RepairJobDesc desc, NodePair nodes, boolean success)
    {
        RepairJob job = syncingJobs.get(desc.columnFamily);
        if (job == null)
        {
            assert terminated;
            return;
        }

        if (!success)
        {
            exception = new RepairException(desc, String.format("Sync failed between %s and %s", nodes.endpoint1, nodes.endpoint2));
            forceShutdown();
            return;
        }

        logger.debug(String.format("[repair #%s] Repair completed between %s and %s on %s", getId(), nodes.endpoint1, nodes.endpoint2, desc.columnFamily));

        if (job.completedSynchronization())
        {
            RepairJob completedJob = syncingJobs.remove(job.desc.columnFamily);
            String remaining = syncingJobs.size() == 0 ? "" : String.format(" (%d remaining column family to sync for this session)", syncingJobs.size());
            if (completedJob != null && completedJob.isFailed())
                logger.warn(String.format("[repair #%s] %s sync failed%s", getId(), desc.columnFamily, remaining));
            else
                logger.info(String.format("[repair #%s] %s is fully synced%s", getId(), desc.columnFamily, remaining));

            if (jobs.isEmpty() && syncingJobs.isEmpty())
            {
                taskExecutor.shutdown();
                // this repair session is completed
                completed.signalAll();
            }
        }
    }

    private String repairedNodes()
    {
        StringBuilder sb = new StringBuilder();
        sb.append(FBUtilities.getBroadcastAddress());
        for (InetAddress ep : endpoints)
            sb.append(", ").append(ep);
        return sb.toString();
    }

    // we don't care about the return value but care about it throwing exception
    public void runMayThrow() throws Exception
    {
        logger.info(String.format("[repair #%s] new session: will sync %s on range %s for %s.%s", getId(), repairedNodes(), range, keyspace, Arrays.toString(cfnames)));

        if (endpoints.isEmpty())
        {
            differencingDone.signalAll();
            logger.info(String.format("[repair #%s] No neighbors to repair with on range %s: session completed", getId(), range));
            return;
        }

        // Checking all nodes are live
        for (InetAddress endpoint : endpoints)
        {
            if (!FailureDetector.instance.isAlive(endpoint))
            {
                String message = String.format("Cannot proceed on repair because a neighbor (%s) is dead: session failed", endpoint);
                differencingDone.signalAll();
                logger.error("[repair #{}] {}", getId(), message);
                throw new IOException(message);
            }
        }

        ActiveRepairService.instance.addToActiveSessions(this);
        try
        {
            // Create and queue a RepairJob for each column family
            for (String cfname : cfnames)
            {
                RepairJob job = new RepairJob(this, parentRepairSession, id, keyspace, cfname, range, parallelismDegree, taskExecutor);
                jobs.offer(job);
            }
            logger.debug("Sending tree requests to endpoints {}", endpoints);
            jobs.peek().sendTreeRequests(endpoints);

            // block whatever thread started this session until all requests have been returned:
            // if this thread dies, the session will still complete in the background
            completed.await();

            if (exception == null)
            {
                logger.info(String.format("[repair #%s] session completed successfully", getId()));
            }
            else
            {
                logger.error(String.format("[repair #%s] session completed with the following error", getId()), exception);
                throw exception;
            }
        }
        catch (InterruptedException e)
        {
            throw new RuntimeException("Interrupted while waiting for repair.");
        }
        finally
        {
            // mark this session as terminated
            terminate();

            ActiveRepairService.instance.removeFromActiveSessions(this);

            // If we've reached here in an exception state without completing Merkle Tree sync, we'll still be registered
            // with the FailureDetector.
            if (fdUnregistered.compareAndSet(false, true))
                FailureDetector.instance.unregisterFailureDetectionEventListener(this);
        }
    }

    public void terminate()
    {
        terminated = true;
        jobs.clear();
        syncingJobs.clear();
    }

    /**
     * clear all RepairJobs and terminate this session.
     */
    public void forceShutdown()
    {
        taskExecutor.shutdownNow();
        differencingDone.signalAll();
        completed.signalAll();
    }

    public void failedSnapshot()
    {
        exception = new IOException("Failed during snapshot creation.");
        forceShutdown();
    }

    void failedNode(InetAddress remote)
    {
        String errorMsg = String.format("Endpoint %s died", remote);
        exception = new IOException(errorMsg);
        // If a node failed during Merkle creation, we stop everything (though there could still be some activity in the background)
        forceShutdown();
    }

    public void onJoin(InetAddress endpoint, EndpointState epState) {}
    public void beforeChange(InetAddress endpoint, EndpointState currentState, ApplicationState newStateKey, VersionedValue newValue) {}
    public void onChange(InetAddress endpoint, ApplicationState state, VersionedValue value) {}
    public void onAlive(InetAddress endpoint, EndpointState state) {}
    public void onDead(InetAddress endpoint, EndpointState state) {}

    public void onRemove(InetAddress endpoint)
    {
        convict(endpoint, Double.MAX_VALUE);
    }

    public void onRestart(InetAddress endpoint, EndpointState epState)
    {
        convict(endpoint, Double.MAX_VALUE);
    }

    public void convict(InetAddress endpoint, double phi)
    {
        if (!endpoints.contains(endpoint))
            return;

        // We want a higher confidence in the failure detection than usual because failing a repair wrongly has a high cost.
        if (phi < 2 * DatabaseDescriptor.getPhiConvictThreshold())
            return;

        // Though unlikely, it is possible to arrive here multiple time and we
        // want to avoid print an error message twice
        if (!isFailed.compareAndSet(false, true))
            return;

        failedNode(endpoint);
    }
}




© 2015 - 2024 Weber Informatics LLC | Privacy Policy