在 flink 集群上远程调试 apache beam 作业

Question

I am running a streaming beam job on a flink cluster where I am getting the following exception.

Caused by: org.apache.beam.sdk.util.UserCodeException: org.apache.flink.streaming.runtime.tasks.ExceptionInChainedOperatorException: Could not forward element to next operator
        at org.apache.beam.sdk.util.UserCodeException.wrap(UserCodeException.java:34)
        at org.apache.beam.sdk.transforms.MapElements$1$DoFnInvoker.invokeProcessElement(Unknown Source)
        at org.apache.beam.runners.core.SimpleDoFnRunner.invokeProcessElement(SimpleDoFnRunner.java:218)
        at org.apache.beam.runners.core.SimpleDoFnRunner.processElement(SimpleDoFnRunner.java:183)
        at org.apache.beam.runners.flink.metrics.DoFnRunnerWithMetricsUpdate.processElement(DoFnRunnerWithMetricsUpdate.java:62)
        at org.apache.beam.runners.flink.translation.wrappers.streaming.DoFnOperator.processElement(DoFnOperator.java:544)
        at org.apache.flink.streaming.runtime.io.StreamInputProcessor.processInput(StreamInputProcessor.java:202)
        at org.apache.flink.streaming.runtime.tasks.OneInputStreamTask.run(OneInputStreamTask.java:105)
        at org.apache.flink.streaming.runtime.tasks.StreamTask.invoke(StreamTask.java:302)
        at org.apache.flink.runtime.taskmanager.Task.run(Task.java:711)
        at java.lang.Thread.run(Thread.java:748)
Caused by: org.apache.flink.streaming.runtime.tasks.ExceptionInChainedOperatorException: Could not forward element to next operator
        at org.apache.flink.streaming.runtime.tasks.OperatorChain$CopyingChainingOutput.pushToOperator(OperatorChain.java:596)
        at org.apache.flink.streaming.runtime.tasks.OperatorChain$CopyingChainingOutput.collect(OperatorChain.java:554)
        at org.apache.flink.streaming.runtime.tasks.OperatorChain$CopyingChainingOutput.collect(OperatorChain.java:534)
        at org.apache.flink.streaming.api.operators.AbstractStreamOperator$CountingOutput.collect(AbstractStreamOperator.java:718)
        at org.apache.flink.streaming.api.operators.AbstractStreamOperator$CountingOutput.collect(AbstractStreamOperator.java:696)
        at org.apache.beam.runners.flink.translation.wrappers.streaming.DoFnOperator$BufferedOutputManager.emit(DoFnOperator.java:941)
        at org.apache.beam.runners.flink.translation.wrappers.streaming.DoFnOperator$BufferedOutputManager.output(DoFnOperator.java:895)
        at org.apache.beam.runners.core.SimpleDoFnRunner.outputWindowedValue(SimpleDoFnRunner.java:252)
        at org.apache.beam.runners.core.SimpleDoFnRunner.access$700(SimpleDoFnRunner.java:74)
        at org.apache.beam.runners.core.SimpleDoFnRunner$DoFnProcessContext.output(SimpleDoFnRunner.java:576)
        at org.apache.beam.sdk.transforms.DoFnOutputReceivers$WindowedContextOutputReceiver.output(DoFnOutputReceivers.java:71)
        at org.apache.beam.sdk.transforms.MapElements$1.processElement(MapElements.java:139)
Caused by: org.apache.beam.sdk.util.UserCodeException: java.lang.IllegalArgumentException: Expect srcResourceIds and destResourceIds have the same scheme, but received alluxio, file.
        at org.apache.beam.sdk.util.UserCodeException.wrap(UserCodeException.java:34)
        at org.apache.beam.sdk.io.WriteFiles$FinalizeTempFileBundles$FinalizeFn$DoFnInvoker.invokeProcessElement(Unknown Source)
        at org.apache.beam.runners.core.SimpleDoFnRunner.invokeProcessElement(SimpleDoFnRunner.java:218)
        at org.apache.beam.runners.core.SimpleDoFnRunner.processElement(SimpleDoFnRunner.java:183)
        at org.apache.beam.runners.flink.metrics.DoFnRunnerWithMetricsUpdate.processElement(DoFnRunnerWithMetricsUpdate.java:62)
        at org.apache.beam.runners.flink.translation.wrappers.streaming.DoFnOperator.processElement(DoFnOperator.java:544)
        at org.apache.flink.streaming.runtime.tasks.OperatorChain$CopyingChainingOutput.pushToOperator(OperatorChain.java:579)
        at org.apache.flink.streaming.runtime.tasks.OperatorChain$CopyingChainingOutput.collect(OperatorChain.java:554)
        at org.apache.flink.streaming.runtime.tasks.OperatorChain$CopyingChainingOutput.collect(OperatorChain.java:534)
        at org.apache.flink.streaming.api.operators.AbstractStreamOperator$CountingOutput.collect(AbstractStreamOperator.java:718)
        at org.apache.flink.streaming.api.operators.AbstractStreamOperator$CountingOutput.collect(AbstractStreamOperator.java:696)
        at org.apache.beam.runners.flink.translation.wrappers.streaming.DoFnOperator$BufferedOutputManager.emit(DoFnOperator.java:941)
        at org.apache.beam.runners.flink.translation.wrappers.streaming.DoFnOperator$BufferedOutputManager.output(DoFnOperator.java:895)
        at org.apache.beam.runners.core.SimpleDoFnRunner.outputWindowedValue(SimpleDoFnRunner.java:252)
        at org.apache.beam.runners.core.SimpleDoFnRunner.access$700(SimpleDoFnRunner.java:74)
        at org.apache.beam.runners.core.SimpleDoFnRunner$DoFnProcessContext.output(SimpleDoFnRunner.java:576)
        at org.apache.beam.sdk.transforms.DoFnOutputReceivers$WindowedContextOutputReceiver.output(DoFnOutputReceivers.java:71)
        at org.apache.beam.sdk.transforms.MapElements$1.processElement(MapElements.java:139)
        at org.apache.beam.sdk.transforms.MapElements$1$DoFnInvoker.invokeProcessElement(Unknown Source)
        at org.apache.beam.runners.core.SimpleDoFnRunner.invokeProcessElement(SimpleDoFnRunner.java:218)
        at org.apache.beam.runners.core.SimpleDoFnRunner.processElement(SimpleDoFnRunner.java:183)
        at org.apache.beam.runners.flink.metrics.DoFnRunnerWithMetricsUpdate.processElement(DoFnRunnerWithMetricsUpdate.java:62)
        at org.apache.beam.runners.flink.translation.wrappers.streaming.DoFnOperator.processElement(DoFnOperator.java:544)
        at org.apache.flink.streaming.runtime.io.StreamInputProcessor.processInput(StreamInputProcessor.java:202)
        at org.apache.flink.streaming.runtime.tasks.OneInputStreamTask.run(OneInputStreamTask.java:105)
        at org.apache.flink.streaming.runtime.tasks.StreamTask.invoke(StreamTask.java:302)
        at org.apache.flink.runtime.taskmanager.Task.run(Task.java:711)
        at java.lang.Thread.run(Thread.java:748)
Caused by: java.lang.IllegalArgumentException: Expect srcResourceIds and destResourceIds have the same scheme, but received alluxio, file.
        at org.apache.beam.vendor.guava.v26_0_jre.com.google.common.base.Preconditions.checkArgument(Preconditions.java:141)
        at org.apache.beam.sdk.io.FileSystems.validateSrcDestLists(FileSystems.java:428)
        at org.apache.beam.sdk.io.FileSystems.rename(FileSystems.java:308)
        at org.apache.beam.sdk.io.FileBasedSink$WriteOperation.moveToOutputFiles(FileBasedSink.java:755)
        at org.apache.beam.sdk.io.WriteFiles$FinalizeTempFileBundles$FinalizeFn.process(WriteFiles.java:850)

The streaming job is getting data from the apache pulsar source and writing output data onto an Alluxio data lake in parquet file format. I am using Spotify's scio for writing this job in Scala. A little code chunk to emphasize what I am trying to achieve:

    pulsarSource
      .open(sc)
      .withFixedWindows(Duration.standardSeconds(windowDuration))
      .toSinkTap(sink)

From the exception, I can see that source and output paths should have the same URI scheme but I don't know how it is happening because I am using an alluxio path as an output directory. There are some temp directories that are being created on alluxio output directory but after the WindowDuration , when the output file is being created, this exception happens. I had a doubt that temp location might be configured by default to the local filesystem, so I did set that to output directory path (alluxio dir path) but it didn't change anything.

sc.options.setTempLocation(outputDir)

I want to do remote debugging in order to figure out the issue. I have tried this document to do remote debugging on the task executor node, but once my IntelliJ IDE connects with the node, I don't get hit on my breakpoint.

Can someone suggest, how can I debug or get more information about this issue. Thanks

Answer 1

Remote debugging might be quite hard, but let's try this first: Make sure you connect to the task manager and not job manager (easy to verify with thread names). Then make sure to have a high number of retries, such that you don't miss the task execution, as attaching the debugging might take a while.

It's also helpful to double check that line numbers of the stack trace match your code version in the IDE. If Flink/Beam is preinstalled, they might run a slightly different version and your break point is void. Just paste the stack trace in your IDE and check if each line matches the expectation. Finally, add a few more breakpoint at central places like org.apache.flink.streaming.runtime.io.StreamInputProcessor.processInput(StreamInputProcessor.java:202) to make sure if the setup is working at all.

However, remote debugging is usually not the recommended option for big data systems. You'd first ensure locally that most of the things work on their own with some IT tests and local runners. Then, you might want to add e2e tests with docker containers and a local mini cluster. Additionally, you'd add a ton of logging statements, which you can turn on and off with your logging configuration. Similarly, if you set logging level to debug, the existing log statements of the frameworks might already be enough to gain some insights. One important thing that you should always look at is the generated topology that you can see in Web UI. Maybe it already tells you the paths in question.