Flink CDC 使用实践以及遇到的问题

背景

最近公司在做一些业务上的架构调整，有一部分是数据从mysql采集到Starrocks，之前的一套方法是走 debezium 到 puslar 到 starrocks,这一套下来比较需要配置很多东西，而且出现问题以后，需要修改很多配置，而且现阶段问题比较多，且采集的是所有线上的数据库，维护起来很费劲。
于是我们进行了采集的数据流调整，使用 Flink CDC这一套，这一套 是端到端的，且采用配置化的方式，支持schema的变更，无需再多一层中间存储层。

最终配置

关于flink cdc的使用配置可以参考Flink CDC 源码解析–整体流程，我能这里只贴出来我们最终使用的配置：

source:type: mysqlname: Database mysql to Data warehousehostname: xxxxport: 3306username: xxxpassword: xxxtables:   db1.table1server-id: 556401-556500scan.startup.mode: initialscan.snapshot.fetch.size: 8096scan.incremental.snapshot.chunk.size: 16192debezium.max.queue.size: 162580debezium.max.batch.size: 40960debezium.poll.interval.ms: 50scan.only.deserialize.captured.tables.changelog.enabled: truescan.parallel-deserialize-changelog.enabled: trueheartbeat.interval: 5sscan.newly-added-table.enabled: truesink:type: starrocksname: StarRocks Sinkjdbc-url: xxxload-url: xxxusername: xxxpassword: xxxsink.buffer-flush.interval-ms: 5000table.create.properties.replication_num: 3table.create.num-buckets: 3route:- source-table: db1.\.*sink-table: db1.<>replace-symbol: <>description: route all tables to starrrockspipeline:name: Sync mysql Database to StarRocksparallelism: 1schema.change.behavior: EVOLVE

遇到的问题

1. EventHeaderV4反序列化问题
   报错如下：

     Caused by: io.debezium.DebeziumException: Failed to deserialize data of EventHeaderV4{timestamp=1732257303000, eventType=WRITE_ROWS, serverId=28555270, headerLength=19, dataLength=320, nextPosition=383299196, flags=0}at io.debezium.connector.mysql.MySqlStreamingChangeEventSource.wrap(MySqlStreamingChangeEventSource.java:1718)... 5 moreCaused by: com.github.shyiko.mysql.binlog.event.deserialization.EventDataDeserializationException: Failed to deserialize data of EventHeaderV4{timestamp=1732257303000, eventType=WRITE_ROWS, serverId=28555270, headerLength=19, dataLength=320, nextPosition=383299196, flags=0}at com.github.shyiko.mysql.binlog.event.deserialization.EventDeserializer.deserializeEventData(EventDeserializer.java:358)at com.github.shyiko.mysql.binlog.event.deserialization.EventDeserializer.nextEvent(EventDeserializer.java:252)at io.debezium.connector.mysql.MySqlStreamingChangeEventSource$1.nextEvent(MySqlStreamingChangeEventSource.java:388)at com.github.shyiko.mysql.binlog.BinaryLogClient.listenForEventPackets(BinaryLogClient.java:1187)... 3 moreCaused by: java.io.EOFException: Failed to read remaining 28 of 36 bytes from position 258280448. Block length: 183. Initial block length: 316.at com.github.shyiko.mysql.binlog.io.ByteArrayInputStream.fill(ByteArrayInputStream.java:115)at com.github.shyiko.mysql.binlog.io.ByteArrayInputStream.read(ByteArrayInputStream.java:105)at io.debezium.connector.mysql.RowDeserializers.deserializeVarString(RowDeserializers.java:264)at io.debezium.connector.mysql.RowDeserializers$WriteRowsDeserializer.deserializeVarString(RowDeserializers.java:192)at com.github.shyiko.mysql.binlog.event.deserialization.AbstractRowsEventDataDeserializer.deserializeCell(AbstractRowsEventDataDeserializer.java:189)at com.github.shyiko.mysql.binlog.event.deserialization.AbstractRowsEventDataDeserializer.deserializeRow(AbstractRowsEventDataDeserializer.java:143)at com.github.shyiko.mysql.binlog.event.deserialization.WriteRowsEventDataDeserializer.deserializeRows(WriteRowsEventDataDeserializer.java:75)at com.github.shyiko.mysql.binlog.event.deserialization.WriteRowsEventDataDeserializer.deserialize(WriteRowsEventDataDeserializer.java:65)at com.github.shyiko.mysql.binlog.event.deserialization.WriteRowsEventDataDeserializer.deserialize(WriteRowsEventDataDeserializer.java:38)at com.github.shyiko.mysql.binlog.event.deserialization.EventDeserializer.deserializeEventData(EventDeserializer.java:352)... 6 more
   

   过段时间自己恢复
   这个现象比较诡异，过段时间就自己恢复了，目前怀疑的点：

   • mysql连接数和带宽问题
   • msyql服务端的配置问题，可以参考Flink CDC FAQ

     mysql> set global slave_net_timeout = 120; 
     mysql> set global thread_pool_idle_timeout = 120;
     

   • 作业反压导致，参考阿里云Flink

     execution.checkpointing.interval=10min
     execution.checkpointing.tolerable-failed-checkpoints=100
     debezium.connect.keep.alive.interval.ms = 40000
     

2. Starrocks Be 内存受限

      java.lang.RuntimeException: com.starrocks.data.load.stream.exception.StreamLoadFailException: Transaction prepare failed, db: shoufuyou_fund, table: fund_common_repay_push, label: flink-4c6c8cfb-5116-4c38-a60e-a1b87cd6f2f2, responseBody: {"Status": "MEM_LIMIT_EXCEEDED","Message": "Memory of process exceed limit. QUERY Backend: 172.17.172.251, fragment: 9048ed6e-6ffb-04db-081b-a4966b179387 Used: 26469550752, Limit: 26316804096. Mem usage has exceed the limit of BE"}errorLog: nullat com.starrocks.data.load.stream.v2.StreamLoadManagerV2.AssertNotException(StreamLoadManagerV2.java:427)at com.starrocks.data.load.stream.v2.StreamLoadManagerV2.write(StreamLoadManagerV2.java:252)at com.starrocks.connector.flink.table.sink.v2.StarRocksWriter.write(StarRocksWriter.java:143)at org.apache.flink.streaming.runtime.operators.sink.SinkWriterOperator.processElement(SinkWriterOperator.java:182)at org.apache.flink.cdc.runtime.operators.sink.DataSinkWriterOperator.processElement(DataSinkWriterOperator.java:178)at org.apache.flink.streaming.runtime.tasks.CopyingChainingOutput.pushToOperator(CopyingChainingOutput.java:75)at org.apache.flink.streaming.runtime.tasks.CopyingChainingOutput.collect(CopyingChainingOutput.java:50)at org.apache.flink.streaming.runtime.tasks.CopyingChainingOutput.collect(CopyingChainingOutput.java:29)at org.apache.flink.streaming.api.operators.StreamMap.processElement(StreamMap.java:38)at org.apache.flink.streaming.runtime.tasks.OneInputStreamTask$StreamTaskNetworkOutput.emitRecord(OneInputStreamTask.java:245)at org.apache.flink.streaming.runtime.io.AbstractStreamTaskNetworkInput.processElement(AbstractStreamTaskNetworkInput.java:217)at org.apache.flink.streaming.runtime.io.AbstractStreamTaskNetworkInput.emitNext(AbstractStreamTaskNetworkInput.java:169)at org.apache.flink.streaming.runtime.io.StreamOneInputProcessor.processInput(StreamOneInputProcessor.java:68)at org.apache.flink.streaming.runtime.tasks.StreamTask.processInput(StreamTask.java:616)at org.apache.flink.streaming.runtime.tasks.mailbox.MailboxProcessor.runMailboxLoop(MailboxProcessor.java:231)at org.apache.flink.streaming.runtime.tasks.StreamTask.runMailboxLoop(StreamTask.java:1071)at org.apache.flink.streaming.runtime.tasks.StreamTask.invoke(StreamTask.java:1020)at org.apache.flink.runtime.taskmanager.Task.runWithSystemExitMonitoring(Task.java:959)at org.apache.flink.runtime.taskmanager.Task.restoreAndInvoke(Task.java:938)at org.apache.flink.runtime.taskmanager.Task.doRun(Task.java:751)at org.apache.flink.runtime.taskmanager.Task.run(Task.java:567)at java.lang.Thread.run(Thread.java:879)Caused by: com.starrocks.data.load.stream.exception.StreamLoadFailException: Transaction prepare failed, db: shoufuyou_fund, table: fund_common_repay_push, label: flink-4c6c8cfb-5116-4c38-a60e-a1b87cd6f2f2, responseBody: {"Status": "MEM_LIMIT_EXCEEDED","Message": "Memory of process exceed limit. QUERY Backend: 172.17.172.251, fragment: 9048ed6e-6ffb-04db-081b-a4966b179387 Used: 26469550752, Limit: 26316804096. Mem usage has exceed the limit of BE"}errorLog: nullat com.starrocks.data.load.stream.TransactionStreamLoader.prepare(TransactionStreamLoader.java:221)at com.starrocks.data.load.stream.v2.TransactionTableRegion.commit(TransactionTableRegion.java:247)at com.starrocks.data.load.stream.v2.StreamLoadManagerV2.lambda$init$0(StreamLoadManagerV2.java:210)... 1 more
   

   由于我们 Starrocks BE的内存是在 32GB，开启多个Flink CDC 任务，会导致CDC初始化的时候，写入BE的数据太多，从而BE内存不够，
   解决： 降低 写入Starrocks的并行读，不要太多CDC同时并行
   也可以参考Troubleshooting StarRocks memory hog issues

3. JobManager Direct buffer memory不够

     java. lang.OutOfMemoryError: Direct buffer memoryat lava.n10.B1ts.reserveMemory(B1ts.lava:/08 ~ 7:1.8.0 312.at java.nio.DirectByteBuffer.<init>(DirectByteBuffer.java:123) ~[7:1.8.0_372]at java.nio.ByteBuffer.allocateDirect(ByteBuffer.java:311) ~[7:1.8.0_3721 at sun.nio.ch.Util.getTemporaryDirectBuffer(Util.java:247) ~[7:1.8.0_372]at sun.nio.ch.IOUtil.write(IOUtil.java:60) ~[7:1.8.0_372]at sun.nio.ch.FileChannelImpl.write(FileChannelImpl.java:234) ~[?:1.8.0_372]at java.nio.channels.Channels.writeFullyImpl(Channels.java:78) ~[?:1.8.0_372]at java.nio.channels.Channels$1.write(Channels.java:174) ~[7:1.8.0_372]at org.apache.flink.core.fs.OffsetAware0utputStream.write(0ffsetAware0utputStream.java:48) ~[ververica-connector-vvp-1.17-vvr-8.0.9-2-SNAPSHOT-jar-with-dependencies.jar:1.17-vvr-8.0.9-2-SNAPSHOT]at org.apache.flink.core.fs.RefCountedFileWithStream.write(RefCountedFileWithStream.java:54) ~[ververica-connector-vvp-1.17-vvr-8.0.9-2-SNAPSHOT-jar-with-dependencies. jar: 1.17-vvr-8.0.9-2-SNAPSHOTat org.apache.flink.core.fs.RefCountedBufferingFileStream.write(RefCountedBufferingFileStream.java:88) ~[ververica-connector-vvp-1.17-vvr-8.0.9-2-SNAPSHOT-jar-with-dependencies.jar:1.17-vvr-8.0.9-2-SNAPSHOT]at ora.aoache.flink.fs.osshadooo.writer.OSSRecoverableFsDataOutoutStream.write OSSRecoverableFsDataOutoutStream.1ava:130) ~?:?at org.apache.flink. runtime.state.filesystem.FsCheckpointMetadata0utputStream.write(FsCheckpointMetadata0utputStream.java:78) ~[flink-dist-1.17-vvr-8.0.9-2-SNAPSHOT. jar:1.17-vvr-8.0.9-2-SNAPSHOT]at java.io.Data0utputStream.write(DataOutputStream.java:107) ~[7:1.8.0_372]at java.io.Filter0utputStream.write(FilterOutputStream.java:97) ~[7:1.8.0_372]at org.apache.flink. runtime.checkpoint.metadata.MetadataV2V3SerializerBase.serializeStreamStateHandle(MetadataV2V3SerializerBase.java:703) ~[flink-dist-1.17-vvr-8.0.9-2-SNAPSHOT. jar:1.17-vvr-8.0.9-2-SNAPSHOT]at org.apache. flink.runtime.checkpoint.metadata.MetadataV3Serializer.serializeStreamStateHandle(MetadataV3Serializer.java:264) ~[flink-dist-1.17-vvr-8.0.9-2-SNAPSHOT.jar:1.17-vvr-8.0.9-2-SNAPSHOT] at org.apache.flink. runtime.checkpoint.metadata.MetadataV3Serializer.serialize0peratorState(MetadataV3Serializer.java:109) ~[flink-dist-1.17-vvr-8.0.9-2-SNAPSHOT.jar:1.17-vvr-8.0.9-2-SNAPSHOT]at org.apache. flink. runtime.checkpoint.metadata.MetadataV2V3SerializerBase.serializeMetadata(MetadataV2V3SerializerBase.java:153) ~[flink-dist-1.17-vvr-8.0.9-2-SNAPSHOT.jar:1.17-vvr-8.0.9-2-SNAPSHOT]at org.apache.flink. runtime.checkpoint.metadata.MetadataV3Serializer.serialize(MetadataV3Serializer.java:83) ~[flink-dist-1.17-vvr-8.0.9-2-SNAPSHOT.jar:1.17-vvr-8.0.9-2-SNAPSHOT] at org.apache.flink.runtime.checkpoint.metadata.MetadataV4Serializer.serialize(MetadataV4Serializer.java:56) ~[flink-dist-1.17-vvr-8.0.9-2-SNAPSHOT.jar:1.17-vvr-8.0.9-2-SNAPSHOT]|at org.apache. flink. runtime.checkpoint.Checkpoints.storeCheckpointMetadata(Checkpoints. java:102) ~[flink-dist-1.17-vvr-8.0.

   解决：
   增加配置:

     jobmanager.memory.off-heap.size: 512mb
   

4. TaskManager jvm内存不够

     java.util.concurrent.TimeoutException: Heartbeat of TaskManager with id job-da2375f5-405b-4398-a568-eaba9711576d-taskmanager-1-34 timed out.at org.apache.flink.runtime.jobmaster.JobMaster$TaskManagerHeartbeatListener.notifyHeartbeatTimeout(JobMaster.java:1714)at org.apache.flink.runtime.heartbeat.DefaultHeartbeatMonitor.run(DefaultHeartbeatMonitor.java:158)at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)at java.util.concurrent.FutureTask.run(FutureTask.java:266)at org.apache.flink.runtime.rpc.akka.AkkaRpcActor.lambda$handleRunAsync$4(AkkaRpcActor.java:453)at org.apache.flink.runtime.concurrent.akka.ClassLoadingUtils.runWithContextClassLoader(ClassLoadingUtils.java:68)at org.apache.flink.runtime.rpc.akka.AkkaRpcActor.handleRunAsync(AkkaRpcActor.java:453)at org.apache.flink.runtime.rpc.akka.AkkaRpcActor.handleRpcMessage(AkkaRpcActor.java:218)at org.apache.flink.runtime.rpc.akka.FencedAkkaRpcActor.handleRpcMessage(FencedAkkaRpcActor.java:84)at org.apache.flink.runtime.rpc.akka.AkkaRpcActor.handleMessage(AkkaRpcActor.java:168)at akka.japi.pf.UnitCaseStatement.apply(CaseStatements.scala:24)at akka.japi.pf.UnitCaseStatement.apply(CaseStatements.scala:20)at scala.PartialFunction.applyOrElse(PartialFunction.scala:127)at scala.PartialFunction.applyOrElse$(PartialFunction.scala:126)at akka.japi.pf.UnitCaseStatement.applyOrElse(CaseStatements.scala:20)at scala.PartialFunction$OrElse.applyOrElse(PartialFunction.scala:175)at scala.PartialFunction$OrElse.applyOrElse(PartialFunction.scala:176)at scala.PartialFunction$OrElse.applyOrElse(PartialFunction.scala:176)at akka.actor.Actor.aroundReceive(Actor.scala:537)at akka.actor.Actor.aroundReceive$(Actor.scala:535)at akka.actor.AbstractActor.aroundReceive(AbstractActor.scala:220)at akka.actor.ActorCell.receiveMessage(ActorCell.scala:579)at akka.actor.ActorCell.invoke(ActorCell.scala:547)at akka.dispatch.Mailbox.processMailbox(Mailbox.scala:270)at akka.dispatch.Mailbox.run(Mailbox.scala:231)at akka.dispatch.Mailbox.exec(Mailbox.scala:243)at java.util.concurrent.ForkJoinTask.doExec(ForkJoinTask.java:289)at java.util.concurrent.ForkJoinPool$WorkQueue.runTask(ForkJoinPool.java:1056)at java.util.concurrent.ForkJoinPool.runWorker(ForkJoinPool.java:1692)at java.util.concurrent.ForkJoinWorkerThread.run(ForkJoinWorkerThread.java:175)

   解决：
   在运行的过程中我们发现TaskManager的 taskmanager.memory.managed.size 内存使用一直为0，这是因为我们这里没有状态的存储，只是ETL,可以参考Flink TaskManager Memory Model
   

   所以增加以下配置

     taskmanager.memory.managed.size: 256mbtaskmanager.memory.process.size: 4096mtable.exec.state.ttl: 1 m
   

5. 读取mysql数据过慢

     java.lang.RuntimeException: One or more fetchers have encountered exceptionat org.apache.flink.connector.base.source.reader.fetcher.SplitFetcherManager.checkErrors(SplitFetcherManager.java:261) ~[flink-connector-files-1.17-vvr-8.0.9-2-SNAPSHOT.jar:1.17-vvr-8.0.9-2-SNAPSHOT]at org.apache.flink.connector.base.source.reader.SourceReaderBase.getNextFetch(SourceReaderBase.java:185) ~[flink-connector-files-1.17-vvr-8.0.9-2-SNAPSHOT.jar:1.17-vvr-8.0.9-2-SNAPSHOT]at org.apache.flink.connector.base.source.reader.SourceReaderBase.pollNext(SourceReaderBase.java:144) ~[flink-connector-files-1.17-vvr-8.0.9-2-SNAPSHOT.jar:1.17-vvr-8.0.9-2-SNAPSHOT]at org.apache.flink.streaming.api.operators.SourceOperator.pollNext(SourceOperator.java:779) ~[flink-dist-1.17-vvr-8.0.9-2-SNAPSHOT.jar:1.17-vvr-8.0.9-2-SNAPSHOT]at org.apache.flink.streaming.api.operators.SourceOperator.emitNext(SourceOperator.java:457) ~[flink-dist-1.17-vvr-8.0.9-2-SNAPSHOT....Caused by: java.lang.RuntimeException: SplitFetcher thread 0 received unexpected exception while polling the recordsat org.apache.flink.connector.base.source.reader.fetcher.SplitFetcher.runOnce(SplitFetcher.java:165) ~[flink-connector-files-1.17-vvr-8.0.9-2-SNAPSHOT.jar:1.17-vvr-8.0.9-2-SNAPSHOT]at org.apache.flink.connector.base.source.reader.fetcher.SplitFetcher.run(SplitFetcher.java:114) ~[flink-connector-files-1.17-vvr-8.0.9-2-SNAPSHOT.jar:1.17-vvr-8.0.9-2-SNAPSHOT]at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511) ~[?:1.8.0_372]at java.util.concurrent.FutureTask.run(FutureTask.java:266) ~[?:1.8.0_372]at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149) ~[?:1.8.0_372]at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624) ~[?:1.8.0_372]... 1 moreCaused by: java.lang.IllegalStateException: The connector is trying to read binlog starting at Struct{version=1.9.8.Final,connector=mysql,name=mysql_binlog_source,ts_ms=1732052840471,db=,server_id=0,file=mysql-bin.051880,pos=347695811,row=0}, but this is no longer available on the server. Reconfigure the connector to use a snapshot when needed.at org.apache.flink.cdc.connectors.mysql.debezium.task.context.StatefulTaskContext.loadStartingOffsetState(StatefulTaskContext.java:212) ~[?:?]at org.apache.flink.cdc.connectors.mysql.debezium.task.context.StatefulTaskContext.configure(StatefulTaskContext.java:133) ~[?:?]at org.apache.flink.cdc.connectors.mysql.debezium.reader.BinlogSplitReader.submitSplit(BinlogSplitReader.java:105) ~[?:?]at org.apache.flink.cdc.connectors.mysql.debezium.reader.BinlogSplitReader.submitSplit(BinlogSplitReader.java:71) ~[?:?]at org.apache.flink.cdc.connectors.mysql.source.reader.MySqlSplitReader.pollSplitRecords(MySqlSplitReader.java:119) ~[?:?]at org.apache.flink.cdc.connectors.mysql.source.reader.MySqlSplitReader.fetch(MySqlSplitReader.java:90) ~[?:?]at org.apache.flink.connector.base.source.reader.fetcher.FetchTask.run(FetchTask.java:58) ~[flink-connector-files-1.17-vvr-8.0.9-2-SNAPSHOT.jar:1.17-vvr-8.0.9-2-SNAPSHOT]at org.apache.flink.connector.base.source.reader.fetcher.SplitFetcher.runOnce(SplitFetcher.java:162) ~[flink-connector-files-1.17-vvr-8.0.9-2-SNAPSHOT.jar:1.17-vvr-8.0.9-2-SNAPSHOT]at org.apache.flink.connector.base.source.reader.fetcher.SplitFetcher.run(SplitFetcher.java:114) ~[flink-connector-files-1.17-vvr-8.0.9-2-SNAPSHOT.jar:1.17-vvr-8.0.9-2-SNAPSHOT]at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511) ~[?:1.8.0_372]at java.util.concurrent.FutureTask.run(FutureTask.java:266) ~[?:1.8.0_372]at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149) ~[?:1.8.0_372]at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624) ~[?:1.8.0_372]... 1 more
   

   解决：
   参考debezium connectors和阿里云,增加如下参数：

   debezium.max.queue.size: 162580
   debezium.max.batch.size: 40960
   debezium.poll.interval.ms: 50
   scan.only.deserialize.captured.tables.changelog.enabled: true
   

6. 增量读取过慢，导致binlog 已经没了
   参考阿里云,增加如下参数

    scan.parallel-deserialize-changelog.enabled: truescan.parallel-deserialize-changelog.handler.size: 4heartbeat.interval: 5s
   

7. 建表不支持 PARTITION BY date_trunc(“week”,date_created)
   在建表中只会出现以下报错信息：

  Unsupported expr ‘date_trunc

可以查看 AnalyzerUtils的1626行：