Invoke teardown when DoFn throws in portable runners #32522

Abacn · 2024-09-20T19:24:35Z

Fix (partly) #31381

Please add a meaningful description for your change here

Thank you for your contribution! Follow this checklist to help us incorporate your contribution quickly and easily:

Mention the appropriate issue in your description (for example: addresses #123), if applicable. This will automatically add a link to the pull request in the issue. If you would like the issue to automatically close on merging the pull request, comment fixes #<ISSUE NUMBER> instead.
Update CHANGES.md with noteworthy changes.
If this contribution is large, please file an Apache Individual Contributor License Agreement.

See the Contributor Guide for more tips on how to make review process smoother.

To check the build health, please visit https://github.com/apache/beam/blob/master/.test-infra/BUILD_STATUS.md

GitHub Actions Tests Status (on master branch)

See CI.md for more information about GitHub Actions CI or the workflows README to see a list of phrases to trigger workflows.

Abacn · 2024-09-20T21:45:24Z

java.lang.RuntimeException: Dataflow job 2024-09-20_12_32_18-8696550461908579531 terminated in state RUNNING but did not return a failure reason.
	at org.apache.beam.runners.dataflow.TestDataflowRunner.run(TestDataflowRunner.java:149)

however the job does succeeded. The worker log shows

2024-09-20 15:34:47.080 EDT
Entering instruction processing loop
2024-09-20 15:35:02.066 EDT
Memory is used/total/max = 392/890/4951 MB, GC last/max = 3.47/3.47 % (configured threshold: 50.00%), #pushbacks=0, gc thrashing=false
2024-09-20 15:40:02.073 EDT
Memory is used/total/max = 392/890/4951 MB, GC last/max = 0.00/3.47 % (configured threshold: 50.00%), #pushbacks=0, gc thrashing=false
2024-09-20 15:45:02.078 EDT
Memory is used/total/max = 88/930/4951 MB, GC last/max = 1.06/3.47 % (configured threshold: 50.00%), #pushbacks=0, gc thrashing=false
2024-09-20 15:47:46.638 EDT
Filepattern gs://apache-beam-samples/shakespeare/sonnets.txt matched 1 files with total size 95662

took extended period of time to match the file source (actually start to process data). Intermittent UW issue?

scwhittle · 2024-09-23T08:23:16Z

sdks/java/harness/src/main/java/org/apache/beam/fn/harness/control/ProcessBundleHandler.java

+        // going to be re-used.
+        for (ThrowingRunnable teardownFunction : Lists.reverse(this.getTearDownFunctions())) {
+          try {
+            teardownFunction.run();


if we throw the exception above during one of the start functions we may not have called all of them. That could be confusing since then finish could be called without corresponding start. If we knew which start/teardown functions corresponeded (ie change to a single list with both matched up) we coudl remember the index we made it to on start functions to call teardown only on the right ones.

The naming here was confusing. startFunction are for startBundle. tearDown corresponds to setup in terms of DoFns, where setup was called as early as:

org.apache.beam.sdk.transforms.reflect.DoFnInvokers.tryInvokeSetupFor(DoFnInvokers.java:53) org.apache.beam.fn.harness.FnApiDoFnRunner.<init>(FnApiDoFnRunner.java:505) org.apache.beam.fn.harness.FnApiDoFnRunner$Factory.createRunnerForPTransform(FnApiDoFnRunner.java:195) org.apache.beam.fn.harness.FnApiDoFnRunner$Factory.createRunnerForPTransform(FnApiDoFnRunner.java:163) org.apache.beam.fn.harness.control.ProcessBundleHandler.createRunnerAndConsumersForPTransformRecursively(ProcessBundleHandler.java:307) org.apache.beam.fn.harness.control.ProcessBundleHandler.createRunnerAndConsumersForPTransformRecursively(ProcessBundleHandler.java:261) org.apache.beam.fn.harness.control.ProcessBundleHandler.createRunnerAndConsumersForPTransformRecursively(ProcessBundleHandler.java:261) org.apache.beam.fn.harness.control.ProcessBundleHandler.createRunnerAndConsumersForPTransformRecursively(ProcessBundleHandler.java:261) org.apache.beam.fn.harness.control.ProcessBundleHandler.createBundleProcessor(ProcessBundleHandler.java:861) org.apache.beam.fn.harness.control.ProcessBundleHandler.lambda$processBundle$0(ProcessBundleHandler.java:511) org.apache.beam.fn.harness.control.ProcessBundleHandler$BundleProcessorCache.get(ProcessBundleHandler.java:972) org.apache.beam.fn.harness.control.ProcessBundleHandler.processBundle(ProcessBundleHandler.java:507)

that was when BundleProcessor gets created. Therefore in bundleProcessor.discard call DoFn's @Setup have been called before.

If the question is about throw the exception in setupFunction, then tearDown is invoked earlier, at

beam/sdks/java/core/src/main/java/org/apache/beam/sdk/transforms/reflect/DoFnInvokers.java

Line 56 in c3be9f0

doFnInvoker.invokeTeardown();

and createBundleProcessor won't complete and discard() method won't invoked

the DoFnInvoker link where teardown is called if exception during startup is for a single DoFn.

If we have fused DoFnA->DoFnB we start in reverse order. If A throws exception in setup, tearDown is invoked on DoFnA. Do we call teardown on B which had startup invoked? If not we probably need some try catch in createBundleProcessor to invoke the added teardown functions.

Tested that teardown on B isn't called, indeed. However this is same for legacy runner, except for legacy runner setup is called in sequence order (not reverse), which means if B throws in setup, tearDown in A won't be called.

Prefer to note this as a separate issue for now.

Abacn · 2024-09-23T15:21:18Z

Java PVR Spark Batch failed twice, each time different test. This is due to #30512. None of them are related to ParDo lifecycle.

Abacn · 2024-09-23T15:28:51Z

Rerun Dataflow Example and test passed. Looks like intermittent issue (cause the SDK harness start to work 12 minutes late)

github-actions · 2024-09-23T15:35:23Z

Checks are failing. Will not request review until checks are succeeding. If you'd like to override that behavior, comment assign set of reviewers

Abacn · 2024-09-24T14:42:38Z

PTAL @scwhittle @robertwb thanks!

github-actions · 2024-09-24T21:05:53Z

Assigning reviewers. If you would like to opt out of this review, comment assign to next reviewer:

R: @kennknowles for label java.
R: @damccorm for label build.

Available commands:

stop reviewer notifications - opt out of the automated review tooling
remind me after tests pass - tag the comment author after tests pass
waiting on author - shift the attention set back to the author (any comment or push by the author will return the attention set to the reviewers)

The PR bot will only process comments in the main thread (not review comments).

sdks/java/harness/src/main/java/org/apache/beam/fn/harness/control/ProcessBundleHandler.java

Abacn · 2024-09-26T15:22:38Z

sdks/java/harness/src/main/java/org/apache/beam/fn/harness/control/ProcessBundleHandler.java

+          } catch (Throwable e) {
+            LOG.warn(
+                "Exceptions are thrown from DoFn.teardown method when trying to discard "
+                    + "ProcessBundleHandler",


adjusted to warning and wording of it. Originally it was an Error, which may cause confuse on the cause of the error in case of a failed pipeline (the cause is upstream error in processElement, etc. not tearDown throws)

github-actions · 2024-10-04T12:14:10Z

Reminder, please take a look at this pr: @kennknowles @damccorm

Abacn · 2024-10-04T17:38:10Z

R: @scwhittle

github-actions · 2024-10-04T17:39:17Z

Stopping reviewer notifications for this pull request: review requested by someone other than the bot, ceding control. If you'd like to restart, comment assign set of reviewers

Abacn · 2024-10-08T21:26:39Z

rebased onto latest HEAD to resolve merge conflict (on CHANGES.md)

scwhittle · 2024-10-09T07:47:41Z

Looks like the spark runner failure might be real from a test that is now not being filtered.

Abacn · 2024-10-09T14:57:27Z

PostCommit Java PVR Spark Batch two attempt different random test failing due to #30512, not related to this PR, merging for now

Abacn force-pushed the invoketeardown branch from 7a0f26b to c0fe7fc Compare September 20, 2024 19:25

github-actions bot added java build runners spark dataflow flink samza labels Sep 20, 2024

Abacn marked this pull request as draft September 20, 2024 21:38

scwhittle reviewed Sep 23, 2024

View reviewed changes

Abacn marked this pull request as ready for review September 23, 2024 15:28

github-actions bot added the Next Action: Reviewers label Sep 24, 2024

scwhittle reviewed Sep 25, 2024

View reviewed changes

sdks/java/harness/src/main/java/org/apache/beam/fn/harness/control/ProcessBundleHandler.java Outdated Show resolved Hide resolved

Abacn commented Sep 26, 2024

View reviewed changes

github-actions bot added the slow-review label Oct 4, 2024

scwhittle approved these changes Oct 7, 2024

View reviewed changes

Abacn added 3 commits October 8, 2024 17:24

Invoke teardown when DoFn throws in portable runners

fd25f7e

update CHANGES.md

1fdd4b8

adjusted comment and logging

413ffe9

Abacn force-pushed the invoketeardown branch from de3433c to 413ffe9 Compare October 8, 2024 21:26

scwhittle approved these changes Oct 9, 2024

View reviewed changes

Abacn merged commit c31d81c into apache:master Oct 9, 2024
30 of 32 checks passed

Abacn deleted the invoketeardown branch October 9, 2024 14:58

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Invoke teardown when DoFn throws in portable runners #32522

Invoke teardown when DoFn throws in portable runners #32522

Abacn commented Sep 20, 2024

Abacn commented Sep 20, 2024 •

edited

Loading

scwhittle Sep 23, 2024

Abacn Sep 23, 2024 •

edited

Loading

Abacn Sep 24, 2024

scwhittle Sep 25, 2024

Abacn Sep 26, 2024 •

edited

Loading

Abacn commented Sep 23, 2024

Abacn commented Sep 23, 2024

github-actions bot commented Sep 23, 2024

Abacn commented Sep 24, 2024

github-actions bot commented Sep 24, 2024

Abacn Sep 26, 2024

github-actions bot commented Oct 4, 2024

Abacn commented Oct 4, 2024

github-actions bot commented Oct 4, 2024

Abacn commented Oct 8, 2024

scwhittle commented Oct 9, 2024

Abacn commented Oct 9, 2024

Invoke teardown when DoFn throws in portable runners #32522

Invoke teardown when DoFn throws in portable runners #32522

Conversation

Abacn commented Sep 20, 2024

GitHub Actions Tests Status (on master branch)

Abacn commented Sep 20, 2024 • edited Loading

scwhittle Sep 23, 2024

Choose a reason for hiding this comment

Abacn Sep 23, 2024 • edited Loading

Choose a reason for hiding this comment

Abacn Sep 24, 2024

Choose a reason for hiding this comment

scwhittle Sep 25, 2024

Choose a reason for hiding this comment

Abacn Sep 26, 2024 • edited Loading

Choose a reason for hiding this comment

Abacn commented Sep 23, 2024

Abacn commented Sep 23, 2024

github-actions bot commented Sep 23, 2024

Abacn commented Sep 24, 2024

github-actions bot commented Sep 24, 2024

Abacn Sep 26, 2024

Choose a reason for hiding this comment

github-actions bot commented Oct 4, 2024

Abacn commented Oct 4, 2024

github-actions bot commented Oct 4, 2024

Abacn commented Oct 8, 2024

scwhittle commented Oct 9, 2024

Abacn commented Oct 9, 2024

Abacn commented Sep 20, 2024 •

edited

Loading

Abacn Sep 23, 2024 •

edited

Loading

Abacn Sep 26, 2024 •

edited

Loading