聊聊maxwell的Recovery

本文主要研究一下maxwell的Recovery

Recovery

maxwell-1.25.1/src/main/java/com/zendesk/maxwell/recovery/Recovery.java

public class Recovery {
    static final Logger LOGGER = LoggerFactory.getLogger(Recovery.class);

    private final ConnectionPool replicationConnectionPool;
    private final RecoveryInfo recoveryInfo;
    private final MaxwellMysqlConfig replicationConfig;
    private final String maxwellDatabaseName;
    private final RecoverySchemaStore schemaStore;


    public Recovery(MaxwellMysqlConfig replicationConfig,
                    String maxwellDatabaseName,
                    ConnectionPool replicationConnectionPool,
                    CaseSensitivity caseSensitivity,
                    RecoveryInfo recoveryInfo) {
        this.replicationConfig = replicationConfig;
        this.replicationConnectionPool = replicationConnectionPool;
        this.recoveryInfo = recoveryInfo;
        this.schemaStore = new RecoverySchemaStore(replicationConnectionPool, maxwellDatabaseName, caseSensitivity);
        this.maxwellDatabaseName = maxwellDatabaseName;
    }

    public HeartbeatRowMap recover() throws Exception {
        String recoveryMsg = String.format(
            "old-server-id: %d, position: %s",
            recoveryInfo.serverID,
            recoveryInfo.position
        );

        LOGGER.warn("attempting to recover from master-change: " + recoveryMsg);
        List<BinlogPosition> list = getBinlogInfo();
        for ( int i = list.size() - 1; i >= 0 ; i-- ) {
            BinlogPosition binlogPosition = list.get(i);
            Position position = Position.valueOf(binlogPosition, recoveryInfo.getHeartbeat());
            Metrics metrics = new NoOpMetrics();

            LOGGER.debug("scanning binlog: " + binlogPosition);
            Replicator replicator = new BinlogConnectorReplicator(
                    this.schemaStore,
                    null,
                    null,
                    replicationConfig,
                    0L, // server-id of 0 activates "mysqlbinlog" behavior where the server will stop after each binlog
                    maxwellDatabaseName,
                    metrics,
                    position,
                    true,
                    recoveryInfo.clientID,
                    new HeartbeatNotifier(),
                    null,
                    new RecoveryFilter(this.maxwellDatabaseName),
                    new MaxwellOutputConfig(),
                    0.25f // Default memory usage size, not used 
            );

            HeartbeatRowMap h = findHeartbeat(replicator);
            if ( h != null ) {
                LOGGER.warn("recovered new master position: " + h.getNextPosition());
                return h;
            }
        }

        LOGGER.error("Could not recover from master-change: " + recoveryMsg);
        return null;
    }

    /**
     * try to find a given heartbeat value from the replicator.
     * @return A BinlogPosition where the heartbeat was found, or null if none was found.
     */
    private HeartbeatRowMap findHeartbeat(Replicator r) throws Exception {
        r.startReplicator();
        for (RowMap row = r.getRow(); row != null ; row = r.getRow()) {
            if (!(row instanceof HeartbeatRowMap)) {
                continue;
            }
            HeartbeatRowMap heartbeatRow = (HeartbeatRowMap) row;
            if (heartbeatRow.getPosition().getLastHeartbeatRead() == recoveryInfo.getHeartbeat())
                return heartbeatRow;
        }
        return null;
    }

    /**
     * fetch a list of binlog positions representing the start of each binlog file
     *
     * @return a list of binlog positions to attempt recovery at
     * */

    private List<BinlogPosition> getBinlogInfo() throws SQLException {
        ArrayList<BinlogPosition> list = new ArrayList<>();
        try ( Connection c = replicationConnectionPool.getConnection() ) {
            ResultSet rs = c.createStatement().executeQuery("SHOW BINARY LOGS");
            while ( rs.next() ) {
                list.add(BinlogPosition.at(4, rs.getString("Log_name")));
            }
        }
        return list;
    }
}
  • Recovery提供了recover方法,它先通过getBinlogInfo方法获取BinlogPosition列表,之后从后往前遍历BinlogPosition构建BinlogConnectorReplicator,然后最后通过findHeartbeat方法查找heartbeatRow.getPosition().getLastHeartbeatRead()为recoveryInfo.getHeartbeat()的HeartbeatRowMap,如果不为null则直接返回

Maxwell

maxwell-1.25.1/src/main/java/com/zendesk/maxwell/Maxwell.java

public class Maxwell implements Runnable {
    protected MaxwellConfig config;
    protected MaxwellContext context;
    protected Replicator replicator;

    static final Logger LOGGER = LoggerFactory.getLogger(Maxwell.class);

    public Maxwell(MaxwellConfig config) throws SQLException, URISyntaxException {
        this(new MaxwellContext(config));
    }

    protected Maxwell(MaxwellContext context) throws SQLException, URISyntaxException {
        this.config = context.getConfig();
        this.context = context;
        this.context.probeConnections();
    }

    //......

    private Position attemptMasterRecovery() throws Exception {
        HeartbeatRowMap recoveredHeartbeat = null;
        MysqlPositionStore positionStore = this.context.getPositionStore();
        RecoveryInfo recoveryInfo = positionStore.getRecoveryInfo(config);

        if ( recoveryInfo != null ) {
            Recovery masterRecovery = new Recovery(
                config.replicationMysql,
                config.databaseName,
                this.context.getReplicationConnectionPool(),
                this.context.getCaseSensitivity(),
                recoveryInfo
            );

            recoveredHeartbeat = masterRecovery.recover();

            if (recoveredHeartbeat != null) {
                // load up the schema from the recovery position and chain it into the
                // new server_id
                MysqlSchemaStore oldServerSchemaStore = new MysqlSchemaStore(
                    context.getMaxwellConnectionPool(),
                    context.getReplicationConnectionPool(),
                    context.getSchemaConnectionPool(),
                    recoveryInfo.serverID,
                    recoveryInfo.position,
                    context.getCaseSensitivity(),
                    config.filter,
                    false
                );

                // Note we associate this schema to the start position of the heartbeat event, so that
                // we pick it up when resuming at the event after the heartbeat.
                oldServerSchemaStore.clone(context.getServerID(), recoveredHeartbeat.getPosition());
                return recoveredHeartbeat.getNextPosition();
            }
        }
        return null;
    }

    //......

}
  • Maxwell的attemptMasterRecovery方法通过positionStore.getRecoveryInfo(config)获取recoveryInfo,若recoveryInfo不为null则创建masterRecovery,执行masterRecovery.recover()获取recoveredHeartbeat,若recoveredHeartbeat不为null则返回recoveredHeartbeat.getNextPosition()

小结

Recovery提供了recover方法,它先通过getBinlogInfo方法获取BinlogPosition列表,之后从后往前遍历BinlogPosition构建BinlogConnectorReplicator,然后最后通过findHeartbeat方法查找heartbeatRow.getPosition().getLastHeartbeatRead()为recoveryInfo.getHeartbeat()的HeartbeatRowMap,如果不为null则直接返回

doc

  • Recovery

原文 

https://segmentfault.com/a/1190000022595107

本站部分文章源于互联网,本着传播知识、有益学习和研究的目的进行的转载,为网友免费提供。如有著作权人或出版方提出异议,本站将立即删除。如果您对文章转载有任何疑问请告之我们,以便我们及时纠正。

PS:推荐一个微信公众号: askHarries 或者qq群:474807195,里面会分享一些资深架构师录制的视频录像:有Spring,MyBatis,Netty源码分析,高并发、高性能、分布式、微服务架构的原理,JVM性能优化这些成为架构师必备的知识体系。还能领取免费的学习资源,目前受益良多

转载请注明原文出处:Harries Blog™ » 聊聊maxwell的Recovery

赞 (0)
分享到:更多 ()

评论 0

  • 昵称 (必填)
  • 邮箱 (必填)
  • 网址