Couchbase Analytics：无结果，Couchbase内部服务器错误，索引带有UNIXTIME毫秒时间戳

当我在时间序列数据上添加类似以下内容的索引时，我遇到了奇怪的分析查询行为：

{
    "clientId": "account-123",
    "orderCode": "49776599-cd7f-458a-b0f2-2d4f28aab85f",
    "timestamp": 1591869354000 // UNIX time milliseconds
}

索引包含所有三个字段

index_clientId_timestamp_orderCode (clientId:string)(timestamp:int64)(orderCode:string)`

古怪的行为

当我过滤

clientId

或

orderCode

如果我在

timestamp

上添加一个过滤器，那么couchbase服务器会出错

在Couchbase Enterprise Edition 6.5.0 build 4960上复制的步骤 在bucket中插入一些数据

插入“数据”（键、值）
值（“key1”，“clientId:“account-123”，“orderCode:“49776599-cd7f-458a-b0f2-2d4f28aab85f”，“timestamp:”1591869354000}），
值（“key2”，“clientId:“account-123”，“orderCode:”8e2a8061-c69e-40ee-8ff4-e1fc6f03f76a”，“timestamp:”1591869355000}），
值（“key3”、{“clientId”：“account-123”、“orderCode”：“757d33a6-85e7-4053-ad8b-bacc69ad0200”、“时间戳”：1591869356000}），
值（“key4”，“clientId:“account-123”，“orderCode:“c6022d17-80bf-497f-8f9d-f865896231dc”，“时间戳”：1591869357000}）

创建分析数据集：

在`data`上创建数据集订单，其中`clientId`=“account-123”；

运行以下分析查询，该查询应按预期工作：

选择“订单”*
从“订单”到“订单”
哪里
`客户ID`='account-123'
和“订单代码”=“49776599-cd7f-458a-b0f2-2d4f28aab85f”
和'timestamp`>1591860000000

创建索引以便查询能够更好地扩展：

在'orders'（'clientId`:STRING，'timestamp`:INT64，'orderCode`:STRING）上创建索引索引\u clientId\u timestamp\u orderCode

再次运行查询：

选择“订单”*
从“订单”到“订单”
哪里
`客户ID`='account-123'
和“订单代码”=“49776599-cd7f-458a-b0f2-2d4f28aab85f”
和'timestamp`>1591860000000

您应该会得到一个错误：

[
  {
    "code": 25000,
    "msg": "Internal error",
    "query_from_user": "SELECT `orders`.*\nFROM `orders` orders\nWHERE\n  `clientId` = 'account-123'\n  AND `orderCode` = '49776599-cd7f-458a-b0f2-2d4f28aab85f'\n  AND `timestamp` > 1591860000000"
  }
]

当我查看

/opt/couchbase/var/lib/couchbase/logs/analytics\u error.log

时，我看到以下错误

2020-06-11T14:58:34.561+00:00 ERRO CBAS.apache.asterix [Executor-104:ClusterController] Unexpected exception
java.lang.NullPointerException: null
        at org.apache.asterix.optimizer.rules.am.BTreeAccessMethod.createKeyVarsAndExprs(BTreeAccessMethod.java:773) ~[asterix-algebra.jar:6.5.0-4960]
        at org.apache.asterix.optimizer.rules.am.BTreeAccessMethod.createIndexSearchPlan(BTreeAccessMethod.java:585) ~[asterix-algebra.jar:6.5.0-4960]
        at org.apache.asterix.optimizer.rules.am.BTreeAccessMethod.applySelectPlanTransformation(BTreeAccessMethod.java:183) ~[asterix-algebra.jar:6.5.0-4960]
        at org.apache.asterix.optimizer.rules.am.IntroduceSelectAccessMethodRule.checkAndApplyTheSelectTransformation(IntroduceSelectAccessMethodRule.java:431) ~[asterix-algebra.jar:6.5.0-4960]
        at org.apache.asterix.optimizer.rules.am.IntroduceSelectAccessMethodRule.checkAndApplyTheSelectTransformation(IntroduceSelectAccessMethodRule.java:343) ~[asterix-algebra.jar:6.5.0-4960]
        at org.apache.asterix.optimizer.rules.am.IntroduceSelectAccessMethodRule.checkAndApplyTheSelectTransformation(IntroduceSelectAccessMethodRule.java:343) ~[asterix-algebra.jar:6.5.0-4960]
        at org.apache.asterix.optimizer.rules.am.IntroduceSelectAccessMethodRule.checkAndApplyTheSelectTransformation(IntroduceSelectAccessMethodRule.java:343) ~[asterix-algebra.jar:6.5.0-4960]
        at org.apache.asterix.optimizer.rules.am.IntroduceSelectAccessMethodRule.rewritePre(IntroduceSelectAccessMethodRule.java:168) ~[asterix-algebra.jar:6.5.0-4960]
        at org.apache.hyracks.algebricks.core.rewriter.base.AbstractRuleController.rewriteOperatorRef(AbstractRuleController.java:86) ~[algebricks-core.jar:6.5.0-4960]
        at org.apache.hyracks.algebricks.compiler.rewriter.rulecontrollers.SequentialFixpointRuleController.rewriteWithRuleCollection(SequentialFixpointRuleController.java:52) ~[algebricks-compiler.jar:6.5.0-4960]
        at org.apache.hyracks.algebricks.core.rewriter.base.HeuristicOptimizer.runOptimizationSets(HeuristicOptimizer.java:89) ~[algebricks-core.jar:6.5.0-4960]
        at org.apache.hyracks.algebricks.core.rewriter.base.HeuristicOptimizer.runLogicalOptimizationSets(HeuristicOptimizer.java:80) ~[algebricks-core.jar:6.5.0-4960]
        at org.apache.hyracks.algebricks.core.rewriter.base.HeuristicOptimizer.optimize(HeuristicOptimizer.java:61) ~[algebricks-core.jar:6.5.0-4960]
        at org.apache.hyracks.algebricks.compiler.api.HeuristicCompilerFactoryBuilder$1$1.optimize(HeuristicCompilerFactoryBuilder.java:94) ~[algebricks-compiler.jar:6.5.0-4960]
        at org.apache.asterix.api.common.APIFramework.compileQuery(APIFramework.java:255) ~[asterix-app.jar:6.5.0-4960]
        at org.apache.asterix.app.translator.QueryTranslator.rewriteCompileQuery(QueryTranslator.java:1982) ~[asterix-app.jar:6.5.0-4960]
        at org.apache.asterix.app.translator.QueryTranslator.lambda$handleQuery$3(QueryTranslator.java:2491) ~[asterix-app.jar:6.5.0-4960]
        at org.apache.asterix.app.translator.QueryTranslator.createAndRunJob(QueryTranslator.java:2623) ~[asterix-app.jar:6.5.0-4960]
        at org.apache.asterix.app.translator.QueryTranslator.deliverResult(QueryTranslator.java:2536) ~[asterix-app.jar:6.5.0-4960]
        at org.apache.asterix.app.translator.QueryTranslator.handleQuery(QueryTranslator.java:2506) ~[asterix-app.jar:6.5.0-4960]
        at org.apache.asterix.app.translator.QueryTranslator.compileAndExecute(QueryTranslator.java:399) ~[asterix-app.jar:6.5.0-4960]
        at org.apache.asterix.app.message.ExecuteStatementRequestMessage.handle(ExecuteStatementRequestMessage.java:153) ~[asterix-app.jar:6.5.0-4960]
        at org.apache.asterix.messaging.CCMessageBroker.receivedMessage(CCMessageBroker.java:64) ~[asterix-app.jar:6.5.0-4960]
        at org.apache.hyracks.control.cc.work.ApplicationMessageWork.lambda$notifyMessageBroker$0(ApplicationMessageWork.java:68) ~[hyracks-control-cc.jar:6.5.0-4960]
        at java.util.concurrent.ThreadPoolExecutor.runWorker(Unknown Source) [?:?]
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(Unknown Source) [?:?]
        at java.lang.Thread.run(Unknown Source) [?:?]

从谓词中删除时间戳筛选器将导致搜索按预期返回。然而，当我为我的雇主在更大的数据集上有效地执行相同的操作时，我根本没有得到匹配项。不过，我无法用上面的例子重现这种行为

---

感谢您的帮助：）

这似乎是优化器中的一个错误

请尝试创建两个索引：

CREATE INDEX index_clientId_orderCode ON `orders`(`clientId`:STRING, `orderCode`:STRING)
CREATE INDEX index_timestamp ON `orders`(`timestamp`:BIGINT)

此外，orderCode值看起来是唯一的（或基数较低）。您是否可以在此基础上创建一个索引，并在查询执行时计算剩余的过滤器

CREATE INDEX index_orderCode ON `orders`(`orderCode`:STRING)

另一个解决方法是保留现有的索引定义，但在时间戳上添加一个上限过滤器：

... AND `timestamp` > 1591860000000 and `timestamp` < 9999999999999999

。。。和'timestamp`>1591860000000和'timestamp`<999999999999

我用最新版本复制了错误：Couchbase Enterprise Edition 6.5.1 build 6299。感谢Dmitry的建议，我将尝试并报告。我一直在使用不同的索引运行基准测试，到目前为止，我想到的最佳组合是在clientId和timestamp上有一个索引，然后在clientId和orderCode上再添加一个。任何其他组合都会导致上述错误，或导致整体性能更差。