[FLINK-19766][table-runtime] Introduce File streaming compaction operators #13744

JingsongLi · 2020-10-22T09:10:49Z

What is the purpose of the change & Brief change log

Introduce Compaction operators:

The compaction operator graph is:
TempFileWriter|parallel ---(InputFile&EndInputFile)---> CompactCoordinator|non-parallel
---(CompactionUnit&EndCompaction)--->CompactOperator|parallel---(PartitionCommitInfo)--->
PartitionCommitter|non-parallel

Because the end message is a kind of barrier of record messages, they can only be transmitted
in the way of full broadcast in the link from coordinator to compact operator.

Introduce CompactCoordinator

This is the single (non-parallel) monitoring task which coordinate input files to compaction units.

Receives in-flight input files inside checkpoint.
Receives all upstream end input messages after the checkpoint completes successfully,
starts coordination.

NOTE: The coordination is a stable algorithm, which can ensure that the downstream can
perform compaction at any time without worrying about fail over.

STATE: This operator stores input files in state, after the checkpoint completes successfully,
input files are taken out from the state for coordination.

Introduce CompactOperator

Receives compaction units to do compaction. Send partition commit information after
compaction finished.

Use BulkFormat to read and use BucketWriter to write.

STATE: This operator stores expired files in state, after the checkpoint completes successfully,
We can ensure that these files will not be used again and they can be deleted from the
file system.

Verifying this change

CompactOperatorsTest
BinPackingTest

Does this pull request potentially affect one of the following parts:

Dependencies (does it add or upgrade a dependency): no
The public API, i.e., is any changed class annotated with @Public(Evolving): no
The serializers: no
The runtime per-record code paths (performance sensitive): no
Anything that affects deployment or recovery: JobManager (and its components), Checkpointing, Kubernetes/Yarn/Mesos, ZooKeeper: no
The S3 file system connector: no

Documentation

Does this pull request introduce a new feature? yes
If yes, how is the feature documented? JavaDocs

flinkbot · 2020-10-22T09:13:15Z

Thanks a lot for your contribution to the Apache Flink project. I'm the @flinkbot. I help the community
to review your pull request. We will use this comment to track the progress of the review.

Automated Checks

Last check on commit 73d3642 (Thu Oct 22 09:13:15 UTC 2020)

Warnings:

No documentation files were touched! Remember to keep the Flink docs up to date!

_{Mention the bot in a comment to re-run the automated checks.}

Review Progress

❓ 1. The [description] looks good.
❓ 2. There is [consensus] that the contribution should go into to Flink.
❓ 3. Needs [attention] from.
❓ 4. The change fits into the overall [architecture].
❓ 5. Overall code [quality] is good.

Please see the Pull Request Review Guide for a full explanation of the review process.

Details

The Bot is tracking the review progress through labels. Labels are applied according to the order of the review items. For consensus, approval by a Flink committer of PMC member is required

Bot commands

The @flinkbot bot supports the following commands:

@flinkbot approve description to approve one or more aspects (aspects: description, consensus, architecture and quality)
@flinkbot approve all to approve all aspects
@flinkbot approve-until architecture to approve everything until architecture
@flinkbot attention @username1 [@username2 ..] to require somebody's attention
@flinkbot disapprove architecture to remove an approval you gave earlier

flinkbot · 2020-10-22T09:42:02Z

CI report:

039f49a Azure: SUCCESS

Bot commands

The @flinkbot bot supports the following commands:

@flinkbot run travis re-run the last Travis build
@flinkbot run azure re-run the last Azure build

lirui-apache · 2020-10-26T08:01:03Z

...me-blink/src/main/java/org/apache/flink/table/filesystem/stream/compact/CompactMessages.java

+	/**
+	 * A partitioned input file.
+	 */
+	public static class InputFile implements CoordinatorInput {


Missing serialVersionUID

lirui-apache · 2020-10-26T08:01:29Z

...me-blink/src/main/java/org/apache/flink/table/filesystem/stream/compact/CompactMessages.java

+	/**
+	 * A flag to end file input.
+	 */
+	public static class EndInputFile implements CoordinatorInput {


Missing serialVersionUID

lirui-apache · 2020-10-26T08:10:30Z

...me-blink/src/main/java/org/apache/flink/table/filesystem/stream/compact/CompactMessages.java

+					.toArray(String[]::new);
+		}
+
+		public boolean isTaskMessage(int taskId) {


What's the purpose of this method?

* {@link CompactionUnit} and {@link EndCompaction} must be sent to the downstream in an orderly * manner, while {@link EndCompaction} is broadcast emitting, so unit and endCompaction use the * broadcast emitting mechanism together. Since unit is broadcast, we want it to be processed by * a single task, so we carry the ID in the unit and let the downstream task select its own unit.

Good catch, there is a bug here, should be:

public boolean isTaskMessage(int taskNumber, int taskId) { return unitId % taskNumber == taskId; }

lirui-apache · 2020-10-26T09:44:32Z

...blink/src/main/java/org/apache/flink/table/filesystem/stream/compact/CompactCoordinator.java

+			if (triggerCommit) {
+				commitUpToCheckpoint(endInputFile.getCheckpointId());
+			}
+		}


Throw an exception for unknown elements?

lirui-apache · 2020-10-26T10:37:58Z

...me-blink/src/main/java/org/apache/flink/table/filesystem/stream/compact/CompactMessages.java

+	}
+
+	/**
+	 * A flag to end file input.


It seems more of a flag to end checkpoint rather than file input?

lirui-apache · 2020-10-26T12:20:30Z

...me-blink/src/main/java/org/apache/flink/table/filesystem/stream/compact/CompactOperator.java

+	}
+
+	@Override
+	public void notifyCheckpointComplete(long checkpointId) throws Exception {


If something goes wrong in this method and the job fails over, will this method be called again for the same checkpointId?

No, then will wait for next checkpoint notify.

lirui-apache · 2020-10-26T12:41:04Z

...k/src/test/java/org/apache/flink/table/filesystem/stream/compact/CompactCoordinatorTest.java

+
+			Assert.assertEquals(7, outputs.size());
+
+			assertUnit(outputs.get(0), 0, "p0", Arrays.asList("f0", "f1", "f4"));


I think CompactCoordinator doesn't guarantee to generate CompactionUnit in any specific order of partitions, right?

You mean the order of partitions? There is no relationship between partitions, so there is no need to guarantee this.

Yeah... but then how could we assert the first output is for p0?

You mean we should sort it before assert?

OK, I'll do it

lirui-apache · 2020-10-27T08:56:05Z

...link/src/test/java/org/apache/flink/table/filesystem/stream/compact/CompactOperatorTest.java

+
+		// check all compacted file generated
+		Assert.assertTrue(fs.exists(new Path(folder, "compacted-f0")));
+		Assert.assertTrue(fs.exists(new Path(folder, "compacted-f2")));


Also verify f3 and f6 are not compacted at this point.

lirui-apache

LGTM

…ators

rmetzger added the review=description? label Oct 22, 2020

rmetzger added component=Connectors/FileSystem component=TableSQL/Runtime labels Oct 22, 2020

lirui-apache reviewed Oct 26, 2020

View reviewed changes

lirui-apache reviewed Oct 27, 2020

View reviewed changes

lirui-apache approved these changes Oct 27, 2020

View reviewed changes

JingsongLi force-pushed the compactOperator branch from c8f3503 to 164c8b5 Compare October 28, 2020 04:13

JingsongLi added 12 commits October 29, 2020 10:41

[FLINK-19766][table-runtime] Introduce File streaming compaction oper…

6ceb696

…ators

checkstyle

081f246

minor fixes

fe114c1

checkstyle

07e3fe5

Address comments

cf39956

Fix testUnitSelection

dbfe01e

Address comment

ef03602

Also verify f3 and f6 are not compacted at this point

34c9585

checkstyle

129d98d

move arguments to context

90ec250

checkstyle

12d05a1

Fix case

039f49a

JingsongLi force-pushed the compactOperator branch from a01ecf0 to 039f49a Compare October 29, 2020 02:41

JingsongLi merged commit 2bbbbc0 into apache:master Oct 29, 2020

JingsongLi deleted the compactOperator branch November 5, 2020 09:38


		Assert.assertEquals(7, outputs.size());

		assertUnit(outputs.get(0), 0, "p0", Arrays.asList("f0", "f1", "f4"));

[FLINK-19766][table-runtime] Introduce File streaming compaction operators #13744

[FLINK-19766][table-runtime] Introduce File streaming compaction operators #13744

Uh oh!

Conversation

JingsongLi commented Oct 22, 2020

What is the purpose of the change & Brief change log

Introduce Compaction operators:

Introduce CompactCoordinator

Introduce CompactOperator

Verifying this change

Does this pull request potentially affect one of the following parts:

Documentation

Uh oh!

flinkbot commented Oct 22, 2020

Automated Checks

Review Progress

Uh oh!

flinkbot commented Oct 22, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

CI report:

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lirui-apache left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

flinkbot commented Oct 22, 2020 •

edited

Loading