当前位置: 首页>>代码示例>>Java>>正文


Java ReadRepairDecision.DC_LOCAL属性代码示例

本文整理汇总了Java中org.apache.cassandra.config.ReadRepairDecision.DC_LOCAL属性的典型用法代码示例。如果您正苦于以下问题:Java ReadRepairDecision.DC_LOCAL属性的具体用法?Java ReadRepairDecision.DC_LOCAL怎么用?Java ReadRepairDecision.DC_LOCAL使用的例子?那么恭喜您, 这里精选的属性代码示例或许可以为您提供帮助。您也可以进一步了解该属性所在org.apache.cassandra.config.ReadRepairDecision的用法示例。


在下文中一共展示了ReadRepairDecision.DC_LOCAL属性的3个代码示例,这些例子默认根据受欢迎程度排序。您可以为喜欢或者感觉有用的代码点赞,您的评价将有助于系统推荐出更棒的Java代码示例。

示例1: getReadExecutor

/**
 * @return an executor appropriate for the configured speculative read policy
 */
public static AbstractReadExecutor getReadExecutor(ReadCommand command, ConsistencyLevel consistencyLevel) throws UnavailableException
{
    Keyspace keyspace = Keyspace.open(command.ksName);
    List<InetAddress> allReplicas = StorageProxy.getLiveSortedEndpoints(keyspace, command.key);
    ReadRepairDecision repairDecision = Schema.instance.getCFMetaData(command.ksName, command.cfName).newReadRepairDecision();
    List<InetAddress> targetReplicas = consistencyLevel.filterForQuery(keyspace, allReplicas, repairDecision);

    // Throw UAE early if we don't have enough replicas.
    consistencyLevel.assureSufficientLiveNodes(keyspace, targetReplicas);

    // Fat client. Speculating read executors need access to cfs metrics and sampled latency, and fat clients
    // can't provide that. So, for now, fat clients will always use NeverSpeculatingReadExecutor.
    if (StorageService.instance.isClientMode())
        return new NeverSpeculatingReadExecutor(command, consistencyLevel, targetReplicas);

    if (repairDecision != ReadRepairDecision.NONE)
        ReadRepairMetrics.attempted.mark();

    ColumnFamilyStore cfs = keyspace.getColumnFamilyStore(command.cfName);
    RetryType retryType = cfs.metadata.getSpeculativeRetry().type;

    // Speculative retry is disabled *OR* there are simply no extra replicas to speculate.
    if (retryType == RetryType.NONE || consistencyLevel.blockFor(keyspace) == allReplicas.size())
        return new NeverSpeculatingReadExecutor(command, consistencyLevel, targetReplicas);

    if (targetReplicas.size() == allReplicas.size())
    {
        // CL.ALL, RRD.GLOBAL or RRD.DC_LOCAL and a single-DC.
        // We are going to contact every node anyway, so ask for 2 full data requests instead of 1, for redundancy
        // (same amount of requests in total, but we turn 1 digest request into a full blown data request).
        return new AlwaysSpeculatingReadExecutor(cfs, command, consistencyLevel, targetReplicas);
    }

    // RRD.NONE or RRD.DC_LOCAL w/ multiple DCs.
    InetAddress extraReplica = allReplicas.get(targetReplicas.size());
    // With repair decision DC_LOCAL all replicas/target replicas may be in different order, so
    // we might have to find a replacement that's not already in targetReplicas.
    if (repairDecision == ReadRepairDecision.DC_LOCAL && targetReplicas.contains(extraReplica))
    {
        for (InetAddress address : allReplicas)
        {
            if (!targetReplicas.contains(address))
            {
                extraReplica = address;
                break;
            }
        }
    }
    targetReplicas.add(extraReplica);

    if (retryType == RetryType.ALWAYS)
        return new AlwaysSpeculatingReadExecutor(cfs, command, consistencyLevel, targetReplicas);
    else // PERCENTILE or CUSTOM.
        return new SpeculatingReadExecutor(cfs, command, consistencyLevel, targetReplicas);
}
 
开发者ID:vcostet,项目名称:cassandra-kmean,代码行数:58,代码来源:AbstractReadExecutor.java

示例2: getReadExecutor

/**
 * @return an executor appropriate for the configured speculative read policy
 */
public static AbstractReadExecutor getReadExecutor(SinglePartitionReadCommand command, ConsistencyLevel consistencyLevel) throws UnavailableException
{
    Keyspace keyspace = Keyspace.open(command.metadata().ksName);
    List<InetAddress> allReplicas = StorageProxy.getLiveSortedEndpoints(keyspace, command.partitionKey());
    // 11980: Excluding EACH_QUORUM reads from potential RR, so that we do not miscount DC responses
    ReadRepairDecision repairDecision = consistencyLevel == ConsistencyLevel.EACH_QUORUM
                                        ? ReadRepairDecision.NONE
                                        : command.metadata().newReadRepairDecision();
    List<InetAddress> targetReplicas = consistencyLevel.filterForQuery(keyspace, allReplicas, repairDecision);

    // Throw UAE early if we don't have enough replicas.
    consistencyLevel.assureSufficientLiveNodes(keyspace, targetReplicas);

    if (repairDecision != ReadRepairDecision.NONE)
    {
        Tracing.trace("Read-repair {}", repairDecision);
        ReadRepairMetrics.attempted.mark();
    }

    ColumnFamilyStore cfs = keyspace.getColumnFamilyStore(command.metadata().cfId);
    SpeculativeRetryParam retry = cfs.metadata.params.speculativeRetry;

    // Speculative retry is disabled *OR* there are simply no extra replicas to speculate.
    // 11980: Disable speculative retry if using EACH_QUORUM in order to prevent miscounting DC responses
    if (retry.equals(SpeculativeRetryParam.NONE)
        || consistencyLevel == ConsistencyLevel.EACH_QUORUM
        || consistencyLevel.blockFor(keyspace) == allReplicas.size())
        return new NeverSpeculatingReadExecutor(keyspace, command, consistencyLevel, targetReplicas);

    if (targetReplicas.size() == allReplicas.size())
    {
        // CL.ALL, RRD.GLOBAL or RRD.DC_LOCAL and a single-DC.
        // We are going to contact every node anyway, so ask for 2 full data requests instead of 1, for redundancy
        // (same amount of requests in total, but we turn 1 digest request into a full blown data request).
        return new AlwaysSpeculatingReadExecutor(keyspace, cfs, command, consistencyLevel, targetReplicas);
    }

    // RRD.NONE or RRD.DC_LOCAL w/ multiple DCs.
    InetAddress extraReplica = allReplicas.get(targetReplicas.size());
    // With repair decision DC_LOCAL all replicas/target replicas may be in different order, so
    // we might have to find a replacement that's not already in targetReplicas.
    if (repairDecision == ReadRepairDecision.DC_LOCAL && targetReplicas.contains(extraReplica))
    {
        for (InetAddress address : allReplicas)
        {
            if (!targetReplicas.contains(address))
            {
                extraReplica = address;
                break;
            }
        }
    }
    targetReplicas.add(extraReplica);

    if (retry.equals(SpeculativeRetryParam.ALWAYS))
        return new AlwaysSpeculatingReadExecutor(keyspace, cfs, command, consistencyLevel, targetReplicas);
    else // PERCENTILE or CUSTOM.
        return new SpeculatingReadExecutor(keyspace, cfs, command, consistencyLevel, targetReplicas);
}
 
开发者ID:scylladb,项目名称:scylla-tools-java,代码行数:62,代码来源:AbstractReadExecutor.java

示例3: getReadExecutor

/**
 * @return an executor appropriate for the configured speculative read policy
 */
public static AbstractReadExecutor getReadExecutor(ReadCommand command, ConsistencyLevel consistencyLevel) throws UnavailableException
{
    Keyspace keyspace = Keyspace.open(command.ksName);
    List<InetAddress> allReplicas = StorageProxy.getLiveSortedEndpoints(keyspace, command.key);
    ReadRepairDecision repairDecision = Schema.instance.getCFMetaData(command.ksName, command.cfName).newReadRepairDecision();
    List<InetAddress> targetReplicas = consistencyLevel.filterForQuery(keyspace, allReplicas, repairDecision);

    // Throw UAE early if we don't have enough replicas.
    consistencyLevel.assureSufficientLiveNodes(keyspace, targetReplicas);

    // Fat client. Speculating read executors need access to cfs metrics and sampled latency, and fat clients
    // can't provide that. So, for now, fat clients will always use NeverSpeculatingReadExecutor.
    if (StorageService.instance.isClientMode())
        return new NeverSpeculatingReadExecutor(command, consistencyLevel, targetReplicas);

    if (repairDecision != ReadRepairDecision.NONE)
        ReadRepairMetrics.attempted.mark();

    ColumnFamilyStore cfs = keyspace.getColumnFamilyStore(command.cfName);
    RetryType retryType = cfs.metadata.getSpeculativeRetry().type;
    
    //@daidong For a single cluster, it will fall into PERCENTILE category.
    //logger.info("@daidong debug: " + "RetryType: " + retryType.name());
    
    // Speculative retry is disabled *OR* there are simply no extra replicas to speculate.
    if (retryType == RetryType.NONE || consistencyLevel.blockFor(keyspace) == allReplicas.size())
        return new NeverSpeculatingReadExecutor(command, consistencyLevel, targetReplicas);

    if (targetReplicas.size() == allReplicas.size())
    {
        // CL.ALL, RRD.GLOBAL or RRD.DC_LOCAL and a single-DC.
        // We are going to contact every node anyway, so ask for 2 full data requests instead of 1, for redundancy
        // (same amount of requests in total, but we turn 1 digest request into a full blown data request).
        return new AlwaysSpeculatingReadExecutor(cfs, command, consistencyLevel, targetReplicas);
    }

    // RRD.NONE or RRD.DC_LOCAL w/ multiple DCs.
    InetAddress extraReplica = allReplicas.get(targetReplicas.size());
    // With repair decision DC_LOCAL all replicas/target replicas may be in different order, so
    // we might have to find a replacement that's not already in targetReplicas.
    if (repairDecision == ReadRepairDecision.DC_LOCAL && targetReplicas.contains(extraReplica))
    {
        for (InetAddress address : allReplicas)
        {
            if (!targetReplicas.contains(address))
            {
                extraReplica = address;
                break;
            }
        }
    }
    targetReplicas.add(extraReplica);

    if (retryType == RetryType.ALWAYS)
        return new AlwaysSpeculatingReadExecutor(cfs, command, consistencyLevel, targetReplicas);
    else // PERCENTILE or CUSTOM. @daidong: return a speculatingReadExecutor.
        return new SpeculatingReadExecutor(cfs, command, consistencyLevel, targetReplicas);
}
 
开发者ID:daidong,项目名称:GraphTrek,代码行数:61,代码来源:AbstractReadExecutor.java


注:本文中的org.apache.cassandra.config.ReadRepairDecision.DC_LOCAL属性示例由纯净天空整理自Github/MSDocs等开源代码及文档管理平台,相关代码片段筛选自各路编程大神贡献的开源项目,源码版权归原作者所有,传播和使用请参考对应项目的License;未经允许,请勿转载。