​Atomikos-XA 事务恢复(3)

Atomikos-XA 事务恢复(3)

Apache ShardingSphere 是一套开源的分布式数据库中间件解决方案组成的生态圈,它由 JDBC、Proxy 和 Sidecar(规划中)这 3 款相互独立,却又能够混合部署配合使用的产品组成。它们均提供标准化的数据分片、分布式事务和数据库治理功能,可适用于如 Java 同构、异构语言、云原生等各种多样化的应用场景。

ShardingSphere 已于2020年4月16日成为 Apache 软件基金会的顶级项目。

Atomikos-XA事务恢复

说事务恢复流程之前,我们来讨论下,会啥会出现事务恢复?XA二阶段提交协议不是强一致性的吗?要解答这个问题,我们就要来看看XA二阶段协议有什么问题?

问题一 :单点故障

由于协调者的重要性,一旦协调者TM发生故障。参与者RM会一直阻塞下去。尤其在第二阶段,协调者发生故障,那么所有的参与者还都处于锁定事务资源的状态中,而无法继续完成事务操作。(如果是协调者挂掉,可以重新选举一个协调者,但是无法解决因为协调者宕机导致的参与者处于阻塞状态的问题)

问题二 :数据不一致

数据不一致。在二阶段提交的阶段二中,当协调者向参与者发送commit请求之后,发生了局部网络异常或者在发送commit请求过程中协调者发生了故障,这回导致只有一部分参与者接受到了commit请求。而在这部分参与者接到commit请求之后就会执行commit操作。但是其他部分未接到commit请求的机器则无法执行事务提交。于是整个分布式系统便出现了数据不一致性的现象。

如何解决?

解决的方案简单,就是我们在事务的操作的每一步,我们都需要对事务状态的日志进行人为的记录,我们可以把日志记录存储在我们想存储的地方,可以是本地存储,也可以中心化的存储。atomikos的开源版本,我们之前也分析了,它是使用内存 + file的方式,存储在本地,这样的话,如果在一个集群系统里面,如果有节点宕机,日志又存储在本地,所以事务不能及时的恢复(需要重启服务)。

Atomikos 多场景下事务恢复。

Atomikos 提供了二种方式,来应对不同场景下的异常情况。

  • 场景一:服务节点不宕机,因为其他的原因,产生需要事务恢复的情况。这个时候才要定时任务进行恢复。
    具体的代码 com.atomikos.icatch.imp.TransactionServiceImp.init() 方法,会初始化一个定时任务,进行事务的恢复。

public synchronized void init ( Properties properties ) throws SysException

    {

    shutdownInProgress_ = false;

    control_ = new com.atomikos.icatch.admin.imp.LogControlImp ( (AdminLog) this.recoveryLog );

        ConfigProperties configProperties = new ConfigProperties(properties);

        long recoveryDelay = configProperties.getRecoveryDelay();

    recoveryTimer = new PooledAlarmTimer(recoveryDelay);

    recoveryTimer.addAlarmTimerListener(new AlarmTimerListener() {

            @Override

            public void alarm(AlarmTimer timer) {

                //进行事务恢复

                performRecovery();

                                         }

        });

    TaskManager.SINGLETON.executeTask(recoveryTimer);

    initialized_ = true;

    }

  • 最终会进入com.atomikos.datasource.xa.XATransactionalResource.recover() 方法。

public void recover() {

    XaResourceRecoveryManager xaResourceRecoveryManager = XaResourceRecoveryManager.getInstance();

    if (xaResourceRecoveryManager != null) { //null for LogCloud recovery

        try {

            xaResourceRecoveryManager.recover(getXAResource());

            } catch (Exception e) {

                refreshXAResource(); //cf case 156968

            }

     }

    }

  • 场景二: 当服务节点宕机重启动过程中进行事务的恢复。具体实现在com.atomikos.datasource.xa.XATransactionalResource.setRecoveryService()方法里面。

@Override

public void setRecoveryService ( RecoveryService recoveryService )

            throws ResourceException

    {

    if ( recoveryService != null ) {

        if ( LOGGER.isTraceEnabled() ) LOGGER.logTrace ( "Installing recovery service on resource "

+ getName () );

        this.branchIdentifier=recoveryService.getName();

    //进行事务恢复

        recover();

    }
}

com.atomikos.datasource.xa.XATransactionalResource.recover() 流程详解。

主代码:

public void recover(XAResource xaResource) throws XAException {

// 根据XA recovery 协议获取 xid

    List<XID> xidsToRecover = retrievePreparedXidsFromXaResource(xaResource);

    Collection<XID> xidsToCommit;

    try {

    // xid 与日志记录的xid进行匹配

        xidsToCommit = retrieveExpiredCommittingXidsFromLog();

        for (XID xid : xidsToRecover) {

            if (xidsToCommit.contains(xid)) {

    //执行 XA commit xid 进行提交

                replayCommit(xid, xaResource);

            } else {

                attemptPresumedAbort(xid, xaResource);

            }

        }

    } catch (LogException couldNotRetrieveCommittingXids) {

        LOGGER.logWarning("Transient error while recovering - will retry later...", couldNotRetrieveCommittingXids);

    }

}

  • 我们来看一下如何根据 XA recovery 协议获取RM端存储的xid。进入方法 retrievePreparedXidsFromXaResource(xaResource), 最后进入 com.atomikos.datasource.xa.RecoveryScan.recoverXids()方法。

public static List<XID> recoverXids(XAResource xaResource, XidSelector selector) throws XAException {

        List<XID> ret = new ArrayList<XID>();

    boolean done = false;

    int flags = XAResource.TMSTARTRSCAN;

    Xid[] xidsFromLastScan = null;

    List<XID> allRecoveredXidsSoFar = new ArrayList<XID>();

    do {

        xidsFromLastScan = xaResource.recover(flags);

        flags = XAResource.TMNOFLAGS;

        done = (xidsFromLastScan == null || xidsFromLastScan.length == 0);

        if (!done) {

// TEMPTATIVELY SET done TO TRUE

// TO TOLERATE ORACLE 8.1.7 INFINITE

// LOOP (ALWAYS RETURNS SAME RECOVER

// SET). IF A NEW SET OF XIDS IS RETURNED

// THEN done WILL BE RESET TO FALSE

        done = true;

        for ( int i = 0; i < xidsFromLastScan.length; i++ ) {

            XID xid = new XID ( xidsFromLastScan[i] );

        // our own XID implements equals and hashCode properly

          if (!allRecoveredXidsSoFar.contains(xid)) {

            // a new xid is returned -> we can not be in a recovery loop -> go on

            allRecoveredXidsSoFar.add(xid);

            done = false;

            if (selector.selects(xid)) {

                ret.add(xid);

            }

          }

        }

       }

    } while (!done);

        return ret;

}

  • 我们重点关注xidsFromLastScan = xaResource.recover(flags); 这个方法,如果我们使用MySQL,那么久会进入 MysqlXAConnection.recover()方法。执行 XA recovery xid 语句来获取 xid。

protected static Xid[] recover(Connection c, int flag) throws XAException {        
/*
         * The XA RECOVER statement returns information for those         XA transactions on the MySQL server that are in the PRE        PARED state. (See Section 13.4.7.2, ???XA
         * Transaction States???.) The output includes a row for e        ach such XA transaction on the server, regardless of wh        ich client started it.
         *
         * XA RECOVER output rows look like this (for an example x        id value consisting of the parts 'abc', 'def', and 7):
         *
         * mysql> XA RECOVER;
         * +----------+--------------+--------------+--------+
         * | formatID | gtrid_length | bqual_length | data |
         * +----------+--------------+--------------+--------+
         * | 7 | 3 | 3 | abcdef |
         * +----------+--------------+--------------+--------+
         *
         * The output columns have the following meanings:
         *
         * formatID is the formatID part of the transaction xid
         * gtrid_length is the length in bytes of the gtrid part
        of the xid
         * bqual_length is the length in bytes of the bqual part
        of the xid
         * data is the concatenation of the gtrid and bqual parts         of the xid
         */
        boolean startRscan = ((flag & TMSTARTRSCAN) > 0);
        boolean endRscan = ((flag & TMENDRSCAN) > 0);
        if (!startRscan && !endRscan && flag != TMNOFLAGS) {
            throw new MysqlXAException(XAException.XAER_INVAL, Messages.getString("MysqlXAConnection.001"), null);
}
//
// We return all recovered XIDs at once, so if not  TMSTARTRSCAN, return no new XIDs
//
// We don't attempt to maintain state to check for TMNOFLAGS "outside" of a scan
//
        if (!startRscan) {
            return new Xid[0];
        }
        ResultSet rs = null;
        Statement stmt = null;
        List<MysqlXid> recoveredXidList = new ArrayList<MysqlXid>();
        try {
        // TODO: Cache this for lifetime of XAConnection
            stmt = c.createStatement();
            rs = stmt.executeQuery("XA RECOVER");
        while (rs.next()) {
            final int formatId = rs.getInt(1);
            int gtridLength = rs.getInt(2);
            int bqualLength = rs.getInt(3);
            byte[] gtridAndBqual = rs.getBytes(4);
            final byte[] gtrid = new byte[gtridLength];
            final byte[] bqual = new byte[bqualLength];
            
            if (gtridAndBqual.length != (gtridLength + bqualLength)) {
                throw new MysqlXAException(XAException.XA_RBPROTO, Messages.getString("MysqlXAConnection.002"), null);
}
            System.arraycopy(gtridAndBqual, 0, gtrid, 0, gtridLength);
            System.arraycopy(gtridAndBqual, gtridLength, bqual, 0, bqualLength);
            recoveredXidList.add(new MysqlXid(gtrid, bqual, formatId));
             }
            } catch (SQLException sqlEx) {
                throw mapXAExceptionFromSQLException(sqlEx);
            } finally {
                if (rs != null) {
                    try {
                        rs.close();
                    } catch (SQLException sqlEx) {
                    throw mapXAExceptionFromSQLException(sqlEx);
                    }
                }
                if (stmt != null) {
                    try {
                        stmt.close();
                    } catch (SQLException sqlEx) {
                        throw mapXAExceptionFromSQLException(sqlEx);
                    }
                }
            }
            int numXids = recoveredXidList.size();
            Xid[] asXids = new Xid[numXids];
           Object[] asObjects = recoveredXidList.toArray();
            for (int i = 0; i < numXids; i++) {
                asXids[i] = (Xid) asObjects[i];
            }
                return asXids;
          }
  • 这里要注意如果Mysql的版本 <5.7.7 ,则不会有任何数据,在以后的版本中Mysql进行了修复,因此如果我们想要使用MySQL充当RM,版本必须 >= 5.7.7 ,原因是:

MySQL 5.6版本在客户端退出的时候,自动把已经prepare的事务回滚了,那么MySQL为什么要这样做?这主要取决于MySQL的内部实现,MySQL 5.7以前的版本,对于prepare的事务,MySQL是不会记录binlog的(官方说是减少fsync,起到了优化的作用)。只有当分布式事务提交的时候才会把前面的操作写入binlog信息,所以对于binlog来说,分布式事务与普通的事务没有区别,而prepare以前的操作信息都保存在连接的IO_CACHE中,如果这个时候客户端退出了,以前的binlog信息都会被丢失,再次重连后允许提交的话,会造成Binlog丢失,从而造成主从数据的不一致,所以官方在客户端退出的时候直接把已经prepare的事务都回滚了!

  • 回到主线再从自己记录的事务日志里面获取XID

Collection<XID> xidsToCommit = retrieveExpiredCommittingXidsFromLog();
我们来看下获取事务日志里面的XID的retrieveExpiredCommittingXidsFromLog()方法。然后进入com.atomikos.recovery.imp.RecoveryLogImp.getCommittingParticipants()方法。

public Collection<ParticipantLogEntry> getCommittingParticipants()

            throws LogReadException {

    Collection<ParticipantLogEntry> committingParticipants = new HashSet<ParticipantLogEntry>();

    Collection<CoordinatorLogEntry> committingCoordinatorLogEntries = repository.findAllCommittingCoordinatorLogEntries();

    for (CoordinatorLogEntry coordinatorLogEntry : committingCoordinatorLogEntries) {

        for (ParticipantLogEntry participantLogEntry : coordinatorLogEntry.participants) {

            committingParticipants.add(participantLogEntry);

        }

    }

    return committingParticipants;

   }

到这里我们来简单介绍一下,事务日志的存储结构。首先是
CoordinatorLogEntry,这是一次XA事务的所有信息实体类。
public class CoordinatorLogEntry implements Serializable {

    //全局事务id

    public final String id;

    

    //是否已经提交

    public final boolean wasCommitted;

    

    /**

         * Only for subtransactions, null otherwise.

         */

    public final String superiorCoordinatorId;

    

    //参与者集合

    public final ParticipantLogEntry[] participants;

    }

再来看一下参与者实体类 ParticipantLogEntry :
public class ParticipantLogEntry implements Serializable {

    private static final long serialVersionUID = 1728296701394899871L;

    /**

         * The ID of the global transaction as known by the transaction core.

         */

    public final String coordinatorId;

/**

     * Identifies the participant within the global transaction.

     */

    public final String uri;

/**

     * When does this participant expire (expressed in millis since Jan 1, 1970)?

     */

    public final long expires;

/**

     * Best-known state of the participant.

     */

    public final TxState state;

/**

     * For diagnostic purposes, null if not relevant.

     */

    public final String resourceName;

}

回到com.atomikos.recovery.xa.DefaultXaRecoveryLog.getExpiredCommittingXids()
方法,可以到获取了一次XA事务过程中,存储的事务日志中的xid。
public Set<XID> getExpiredCommittingXids() throws LogReadException {

    Set<XID> ret = new HashSet<XID>();

    Collection<ParticipantLogEntry> entries = log.getCommittingParticipants();

    for (ParticipantLogEntry entry : entries) {

        if (expired(entry) && !http(entry)) {

            XID xid = new XID(entry.coordinatorId, entry.uri);

            ret.add(xid);

        }

       }

        return ret;

      }

如果从RM中通过XA recovery取出的XID,包含在从事务日志中取出的XID,
则进行commit,否则进行rollback.
List<XID> xidsToRecover = retrievePreparedXidsFromXaResource(xaResource);

    Collection<XID> xidsToCommit;

    try {

        xidsToCommit = retrieveExpiredCommittingXidsFromLog();

        for (XID xid : xidsToRecover) {

            if (xidsToCommit.contains(xid)) {

                replayCommit(xid, xaResource);

            } else {

                attemptPresumedAbort(xid, xaResource);

            }

           }

          } catch (LogException couldNotRetrieveCommittingXids) {

            LOGGER.logWarning("Transient error while recovering - will retry later...", couldNotRetrieveCommittingXids);

}

replayCommit 方法如下:
private void replayCommit(XID xid, XAResource xaResource) {

    if (LOGGER.isDebugEnabled()) LOGGER.logDebug("Replaying commit of xid: " + xid);

    try {

    //进行事务提交

        xaResource.commit(xid, false);

    //更新事务日志

        log.terminated(xid);

    } catch (XAException e) {

        if (alreadyHeuristicallyTerminatedByResource(e)) {

            handleHeuristicTerminationByResource(xid, xaResource, e, true);

        } else if (xidTerminatedInResourceByConcurrentCommit(e)) {

            log.terminated(xid);

        } else {

            LOGGER.logWarning("Transient error while replaying commit - will retry later...", e);

        }

       }

      }
attemptPresumedAbort(xid, xaResource); 方法如下:
private void attemptPresumedAbort(XID xid, XAResource xaResource) {

    try {

        log.presumedAborting(xid);

        if (LOGGER.isDebugEnabled()) LOGGER.logDebug("Presumed abort of xid: " + xid);

        try {

        //进行回滚

            xaResource.rollback(xid);

        //更新日志状态

            log.terminated(xid);

        } catch (XAException e) {

            if (alreadyHeuristicallyTerminatedByResource(e)) {

                handleHeuristicTerminationByResource(xid, xaResource, e, false);

            } else if (xidTerminatedInResourceByConcurrentRollback(e)) {

            log.terminated(xid);

           } else {

            LOGGER.logWarning("Unexpected exception during recovery - ignoring to retry later...", e);

           }

        }

    } catch (IllegalStateException presumedAbortNotAllowedInCurrentLogState) {

        // ignore to retry later if necessary

    } catch (LogException logWriteException) {

        LOGGER.logWarning("log write failed for Xid: "+xid+", ignoring to retry later", logWriteException);

        }

    }

文章到此,已经写的很长很多了,我们分析了ShardingSphere对于XA方案,提供了一套SPI解决方案,对Atomikos进行了整合,也分析了Atomikos初始化流程,开始事务流程,获取连接流程,提交事务流程,回滚事务流程,事务恢复流程。
希望对大家理解XA的原理有所帮助。

关于我们

 Apache ShardingSphere是一套开源的分布式数据库中间件解决方案组成的生态圈,它由Sharding-JDBC、Sharding-Proxy和Sharding-Sidecar(规划中)这3款相互独立的产品组成。他们均提供标准化的数据分片、分布式事务、数据迁移、数据库治理和管控界面功能,可适用于如Java同构、异构语言、容器、云原生等各种多样化的应用场景。

Apache ShardingSphere不断践行Apache Way,致力于打造充满活力、规范、互助的社区!开源路上,我们欢迎你的加入。

项目地址:

https://github.com/apache/shardingsphere

更多信息请浏览官网

https://shardingsphere.apache.org/

作者介绍:肖宇,Apache ShardingSphere Committer,开源hmily分布式事务框架作者,
开源soul网关作者,热爱开源,追求写优雅代码。目前就职于京东数科,参与ShardingSphere的开源建设,以及分布式数据库的研发工作。

扫码关注

ShardingSphere

评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值