You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@flink.apache.org by dm...@apache.org on 2023/03/15 15:59:23 UTC
[flink] 02/02: [FLINK-31317] Introduce JobResourceRequirements and JobVertexResourceRequirements data structures.
This is an automated email from the ASF dual-hosted git repository.
dmvk pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/flink.git
commit a84de2ca4a9239bccb0d28c9bb6f841de1879bf9
Author: David Moravek <dm...@apache.org>
AuthorDate: Mon Feb 27 18:48:28 2023 +0100
[FLINK-31317] Introduce JobResourceRequirements and JobVertexResourceRequirements data structures.
Signed-off-by: David Moravek <dm...@apache.org>
---
.../runtime/jobgraph/JobResourceRequirements.java | 235 +++++++++++++++++++++
.../jobgraph/JobVertexResourceRequirements.java | 119 +++++++++++
.../jobgraph/JobResourceRequirementsTest.java | 175 +++++++++++++++
3 files changed, 529 insertions(+)
diff --git a/flink-runtime/src/main/java/org/apache/flink/runtime/jobgraph/JobResourceRequirements.java b/flink-runtime/src/main/java/org/apache/flink/runtime/jobgraph/JobResourceRequirements.java
new file mode 100644
index 00000000000..47d99709a93
--- /dev/null
+++ b/flink-runtime/src/main/java/org/apache/flink/runtime/jobgraph/JobResourceRequirements.java
@@ -0,0 +1,235 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.flink.runtime.jobgraph;
+
+import org.apache.flink.util.InstantiationUtil;
+
+import java.io.IOException;
+import java.io.Serializable;
+import java.util.ArrayList;
+import java.util.Collections;
+import java.util.HashMap;
+import java.util.HashSet;
+import java.util.List;
+import java.util.Map;
+import java.util.Objects;
+import java.util.Optional;
+import java.util.Set;
+
+import static org.apache.flink.util.Preconditions.checkNotNull;
+
+/** Information about the parallelism of job vertices. */
+public class JobResourceRequirements implements Serializable {
+
+ private static final long serialVersionUID = 1L;
+
+ /**
+ * A key for an internal config option (intentionally prefixed with $internal to make this
+ * explicit), that we'll serialize the {@link JobResourceRequirements} into, when writing it to
+ * {@link JobGraph}.
+ */
+ private static final String JOB_RESOURCE_REQUIREMENTS_KEY =
+ "$internal.job-resource-requirements";
+
+ private static final JobResourceRequirements EMPTY =
+ new JobResourceRequirements(Collections.emptyMap());
+
+ /**
+ * Write {@link JobResourceRequirements resource requirements} into the configuration of a given
+ * {@link JobGraph}.
+ *
+ * @param jobGraph job graph to write requirements to
+ * @param jobResourceRequirements resource requirements to write
+ * @throws IOException in case we're not able to serialize requirements into the configuration
+ */
+ public static void writeToJobGraph(
+ JobGraph jobGraph, JobResourceRequirements jobResourceRequirements) throws IOException {
+ InstantiationUtil.writeObjectToConfig(
+ jobResourceRequirements,
+ jobGraph.getJobConfiguration(),
+ JOB_RESOURCE_REQUIREMENTS_KEY);
+ }
+
+ /**
+ * Read {@link JobResourceRequirements resource requirements} from the configuration of a given
+ * {@link JobGraph}.
+ *
+ * @param jobGraph job graph to read requirements from
+ * @throws IOException in case we're not able to deserialize requirements from the configuration
+ */
+ public static Optional<JobResourceRequirements> readFromJobGraph(JobGraph jobGraph)
+ throws IOException {
+ try {
+ return Optional.ofNullable(
+ InstantiationUtil.readObjectFromConfig(
+ jobGraph.getJobConfiguration(),
+ JOB_RESOURCE_REQUIREMENTS_KEY,
+ JobResourceRequirements.class.getClassLoader()));
+ } catch (ClassNotFoundException e) {
+ throw new IOException(
+ "Unable to deserialize JobResourceRequirements due to missing classes. This might happen when the JobGraph was written from a different Flink version.",
+ e);
+ }
+ }
+
+ /**
+ * This method validates that:
+ *
+ * <ul>
+ * <li>The requested boundaries are less or equal than the max parallelism.
+ * <li>The requested boundaries are greater than zero.
+ * <li>The requested upper bound is greater than the lower bound.
+ * <li>There are no unknown job vertex ids and that we're not missing any.
+ * </ul>
+ *
+ * In case any boundary is set to {@code -1}, it will be expanded to the default value ({@code
+ * 1} for the lower bound and the max parallelism for the upper bound), before the validation.
+ *
+ * @param jobResourceRequirements contains the new resources requirements for the job vertices
+ * @param maxParallelismPerVertex allows us to look up maximum possible parallelism for a job
+ * vertex
+ * @return a list of validation errors
+ */
+ public static List<String> validate(
+ JobResourceRequirements jobResourceRequirements,
+ Map<JobVertexID, Integer> maxParallelismPerVertex) {
+ final List<String> errors = new ArrayList<>();
+ final Set<JobVertexID> missingJobVertexIds =
+ new HashSet<>(maxParallelismPerVertex.keySet());
+ for (JobVertexID jobVertexId : jobResourceRequirements.getJobVertices()) {
+ missingJobVertexIds.remove(jobVertexId);
+ final Optional<Integer> maybeMaxParallelism =
+ Optional.ofNullable(maxParallelismPerVertex.get(jobVertexId));
+ if (maybeMaxParallelism.isPresent()) {
+ final JobVertexResourceRequirements.Parallelism requestedParallelism =
+ jobResourceRequirements.findParallelism(jobVertexId).get();
+ int lowerBound =
+ requestedParallelism.getLowerBound() == -1
+ ? 1
+ : requestedParallelism.getLowerBound();
+ int upperBound =
+ requestedParallelism.getUpperBound() == -1
+ ? maybeMaxParallelism.get()
+ : requestedParallelism.getUpperBound();
+ if (lowerBound < 1 || upperBound < 1) {
+ errors.add(
+ String.format(
+ "Both, the requested lower bound [%d] and upper bound [%d] for job vertex [%s] must be greater than zero.",
+ lowerBound, upperBound, jobVertexId));
+ // Don't validate this vertex any further to avoid additional noise.
+ continue;
+ }
+ if (lowerBound > upperBound) {
+ errors.add(
+ String.format(
+ "The requested lower bound [%d] for job vertex [%s] is higher than the upper bound [%d].",
+ lowerBound, jobVertexId, upperBound));
+ }
+ if (maybeMaxParallelism.get() < upperBound) {
+ errors.add(
+ String.format(
+ "The newly requested parallelism %d for the job vertex %s exceeds its maximum parallelism %d.",
+ upperBound, jobVertexId, maybeMaxParallelism.get()));
+ }
+ } else {
+ errors.add(
+ String.format(
+ "Job vertex [%s] was not found in the JobGraph.", jobVertexId));
+ }
+ }
+ for (JobVertexID jobVertexId : missingJobVertexIds) {
+ errors.add(
+ String.format(
+ "The request is incomplete, missing job vertex [%s] resource requirements.",
+ jobVertexId));
+ }
+ return errors;
+ }
+
+ public static JobResourceRequirements empty() {
+ return JobResourceRequirements.EMPTY;
+ }
+
+ public static Builder newBuilder() {
+ return new Builder();
+ }
+
+ public static final class Builder {
+
+ private final Map<JobVertexID, JobVertexResourceRequirements> vertexResources =
+ new HashMap<>();
+
+ public Builder setParallelismForJobVertex(
+ JobVertexID jobVertexId, int lowerBound, int upperBound) {
+ vertexResources.put(
+ jobVertexId,
+ new JobVertexResourceRequirements(
+ new JobVertexResourceRequirements.Parallelism(lowerBound, upperBound)));
+ return this;
+ }
+
+ public JobResourceRequirements build() {
+ return new JobResourceRequirements(vertexResources);
+ }
+ }
+
+ private final Map<JobVertexID, JobVertexResourceRequirements> vertexResources;
+
+ public JobResourceRequirements(
+ Map<JobVertexID, JobVertexResourceRequirements> vertexResources) {
+ this.vertexResources =
+ Collections.unmodifiableMap(new HashMap<>(checkNotNull(vertexResources)));
+ }
+
+ public Optional<JobVertexResourceRequirements.Parallelism> findParallelism(
+ JobVertexID jobVertexId) {
+ return Optional.ofNullable(vertexResources.get(jobVertexId))
+ .map(JobVertexResourceRequirements::getParallelism);
+ }
+
+ public Set<JobVertexID> getJobVertices() {
+ return vertexResources.keySet();
+ }
+
+ public Map<JobVertexID, JobVertexResourceRequirements> getJobVertexParallelisms() {
+ return vertexResources;
+ }
+
+ @Override
+ public boolean equals(Object o) {
+ if (this == o) {
+ return true;
+ }
+ if (o == null || getClass() != o.getClass()) {
+ return false;
+ }
+ final JobResourceRequirements that = (JobResourceRequirements) o;
+ return Objects.equals(vertexResources, that.vertexResources);
+ }
+
+ @Override
+ public int hashCode() {
+ return Objects.hash(vertexResources);
+ }
+
+ @Override
+ public String toString() {
+ return "JobResourceRequirements{" + "vertexResources=" + vertexResources + '}';
+ }
+}
diff --git a/flink-runtime/src/main/java/org/apache/flink/runtime/jobgraph/JobVertexResourceRequirements.java b/flink-runtime/src/main/java/org/apache/flink/runtime/jobgraph/JobVertexResourceRequirements.java
new file mode 100644
index 00000000000..3930501b002
--- /dev/null
+++ b/flink-runtime/src/main/java/org/apache/flink/runtime/jobgraph/JobVertexResourceRequirements.java
@@ -0,0 +1,119 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.flink.runtime.jobgraph;
+
+import org.apache.flink.shaded.jackson2.com.fasterxml.jackson.annotation.JsonCreator;
+import org.apache.flink.shaded.jackson2.com.fasterxml.jackson.annotation.JsonProperty;
+
+import java.io.Serializable;
+import java.util.Objects;
+
+import static org.apache.flink.util.Preconditions.checkNotNull;
+
+/** Information about the parallelism of job vertices. */
+public class JobVertexResourceRequirements implements Serializable {
+
+ private static final long serialVersionUID = 1L;
+
+ private static final String FIELD_NAME_PARALLELISM = "parallelism";
+
+ public static class Parallelism implements Serializable {
+
+ private static final String FIELD_NAME_LOWER_BOUND = "lowerBound";
+ private static final String FIELD_NAME_UPPER_BOUND = "upperBound";
+
+ @JsonProperty(FIELD_NAME_LOWER_BOUND)
+ private final int lowerBound;
+
+ @JsonProperty(FIELD_NAME_UPPER_BOUND)
+ private final int upperBound;
+
+ @JsonCreator
+ public Parallelism(
+ @JsonProperty(FIELD_NAME_LOWER_BOUND) int lowerBound,
+ @JsonProperty(FIELD_NAME_UPPER_BOUND) int upperBound) {
+ this.lowerBound = lowerBound;
+ this.upperBound = upperBound;
+ }
+
+ public int getLowerBound() {
+ return lowerBound;
+ }
+
+ public int getUpperBound() {
+ return upperBound;
+ }
+
+ @Override
+ public boolean equals(Object o) {
+ if (this == o) {
+ return true;
+ }
+ if (o == null || getClass() != o.getClass()) {
+ return false;
+ }
+ final Parallelism that = (Parallelism) o;
+ return lowerBound == that.lowerBound && upperBound == that.upperBound;
+ }
+
+ @Override
+ public int hashCode() {
+ return Objects.hash(lowerBound, upperBound);
+ }
+
+ @Override
+ public String toString() {
+ return "Parallelism{" + "lowerBound=" + lowerBound + ", upperBound=" + upperBound + '}';
+ }
+ }
+
+ @JsonProperty(FIELD_NAME_PARALLELISM)
+ private final Parallelism parallelism;
+
+ public JobVertexResourceRequirements(
+ @JsonProperty(FIELD_NAME_PARALLELISM) Parallelism parallelism) {
+ this.parallelism = checkNotNull(parallelism);
+ }
+
+ public Parallelism getParallelism() {
+ return parallelism;
+ }
+
+ @Override
+ public boolean equals(Object o) {
+ if (this == o) {
+ return true;
+ }
+ if (o == null || getClass() != o.getClass()) {
+ return false;
+ }
+ final JobVertexResourceRequirements that = (JobVertexResourceRequirements) o;
+ return parallelism.equals(that.parallelism);
+ }
+
+ @Override
+ public int hashCode() {
+ return Objects.hash(parallelism);
+ }
+
+ @Override
+ public String toString() {
+ return "JobVertexResourceRequirements{" + "parallelism=" + parallelism + '}';
+ }
+}
diff --git a/flink-runtime/src/test/java/org/apache/flink/runtime/jobgraph/JobResourceRequirementsTest.java b/flink-runtime/src/test/java/org/apache/flink/runtime/jobgraph/JobResourceRequirementsTest.java
new file mode 100644
index 00000000000..db39062420f
--- /dev/null
+++ b/flink-runtime/src/test/java/org/apache/flink/runtime/jobgraph/JobResourceRequirementsTest.java
@@ -0,0 +1,175 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.flink.runtime.jobgraph;
+
+import org.apache.flink.util.TestLoggerExtension;
+
+import org.junit.jupiter.api.Test;
+import org.junit.jupiter.api.extension.ExtendWith;
+
+import java.io.IOException;
+import java.util.Collections;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+import static org.assertj.core.api.Assertions.assertThat;
+
+/** Tests for {@link org.apache.flink.runtime.jobgraph.JobResourceRequirements}. */
+@ExtendWith(TestLoggerExtension.class)
+class JobResourceRequirementsTest {
+
+ private final JobVertexID firstVertexId = new JobVertexID();
+ private final JobVertexID secondVertexId = new JobVertexID();
+
+ @Test
+ void testSuccessfulValidation() {
+ final JobResourceRequirements jobResourceRequirements =
+ JobResourceRequirements.newBuilder()
+ .setParallelismForJobVertex(firstVertexId, 1, 4)
+ .setParallelismForJobVertex(secondVertexId, 1, 4)
+ .build();
+ final Map<JobVertexID, Integer> maxParallelismPerVertex = new HashMap<>();
+ maxParallelismPerVertex.put(firstVertexId, 10);
+ maxParallelismPerVertex.put(secondVertexId, 10);
+ final List<String> validationErrors =
+ JobResourceRequirements.validate(jobResourceRequirements, maxParallelismPerVertex);
+ assertThat(validationErrors).isEmpty();
+ }
+
+ @Test
+ void testValidateVertexIdsNotFoundInJobGraph() {
+ final JobResourceRequirements jobResourceRequirements =
+ JobResourceRequirements.newBuilder()
+ .setParallelismForJobVertex(firstVertexId, 1, 4)
+ .setParallelismForJobVertex(secondVertexId, 1, 4)
+ .build();
+ final List<String> validationErrors =
+ JobResourceRequirements.validate(jobResourceRequirements, Collections.emptyMap());
+ assertThat(validationErrors).hasSize(2);
+ for (String validationError : validationErrors) {
+ assertThat(validationError).contains("was not found in the JobGraph");
+ }
+ }
+
+ @Test
+ void testValidateUpperBoundHigherThanMaxParallelism() {
+ final JobResourceRequirements jobResourceRequirements =
+ JobResourceRequirements.newBuilder()
+ .setParallelismForJobVertex(firstVertexId, 1, 10)
+ .setParallelismForJobVertex(secondVertexId, 1, 5)
+ .build();
+ final Map<JobVertexID, Integer> maxParallelismPerVertex = new HashMap<>();
+ maxParallelismPerVertex.put(firstVertexId, 5);
+ maxParallelismPerVertex.put(secondVertexId, 5);
+ final List<String> validationErrors =
+ JobResourceRequirements.validate(jobResourceRequirements, maxParallelismPerVertex);
+ assertThat(validationErrors).hasSize(1);
+ for (String validationError : validationErrors) {
+ assertThat(validationError).contains("exceeds its maximum parallelism");
+ }
+ }
+
+ @Test
+ void testValidateIncompleteRequirements() {
+ final JobResourceRequirements jobResourceRequirements =
+ JobResourceRequirements.newBuilder()
+ .setParallelismForJobVertex(firstVertexId, 1, 10)
+ .build();
+ final Map<JobVertexID, Integer> maxParallelismPerVertex = new HashMap<>();
+ maxParallelismPerVertex.put(firstVertexId, 10);
+ maxParallelismPerVertex.put(secondVertexId, 10);
+ final List<String> validationErrors =
+ JobResourceRequirements.validate(jobResourceRequirements, maxParallelismPerVertex);
+ assertThat(validationErrors).hasSize(1);
+ for (String validationError : validationErrors) {
+ assertThat(validationError).contains("request is incomplete");
+ }
+ }
+
+ @Test
+ void testValidateLowerBoundDoesNotExceedUpperBound() {
+ final JobResourceRequirements jobResourceRequirements =
+ JobResourceRequirements.newBuilder()
+ .setParallelismForJobVertex(firstVertexId, 10, 9)
+ .setParallelismForJobVertex(secondVertexId, 10, 10)
+ .build();
+ final Map<JobVertexID, Integer> maxParallelismPerVertex = new HashMap<>();
+ maxParallelismPerVertex.put(firstVertexId, 10);
+ maxParallelismPerVertex.put(secondVertexId, 10);
+ final List<String> validationErrors =
+ JobResourceRequirements.validate(jobResourceRequirements, maxParallelismPerVertex);
+ assertThat(validationErrors).hasSize(1);
+ for (String validationError : validationErrors) {
+ assertThat(validationError).contains("is higher than the upper bound");
+ }
+ }
+
+ @Test
+ void testValidateLowerOrUpperBoundIsLowerThanOne() {
+ final JobResourceRequirements jobResourceRequirements =
+ JobResourceRequirements.newBuilder()
+ .setParallelismForJobVertex(firstVertexId, 0, 10)
+ .setParallelismForJobVertex(secondVertexId, 1, 0)
+ .build();
+ final Map<JobVertexID, Integer> maxParallelismPerVertex = new HashMap<>();
+ maxParallelismPerVertex.put(firstVertexId, 10);
+ maxParallelismPerVertex.put(secondVertexId, 10);
+ final List<String> validationErrors =
+ JobResourceRequirements.validate(jobResourceRequirements, maxParallelismPerVertex);
+ assertThat(validationErrors).hasSize(2);
+ for (String validationError : validationErrors) {
+ assertThat(validationError).contains("must be greater than zero");
+ }
+ }
+
+ @Test
+ void testValidateDefaults() {
+ final JobResourceRequirements jobResourceRequirements =
+ JobResourceRequirements.newBuilder()
+ .setParallelismForJobVertex(firstVertexId, -1, -1)
+ .build();
+ final Map<JobVertexID, Integer> maxParallelismPerVertex = new HashMap<>();
+ maxParallelismPerVertex.put(firstVertexId, 10);
+ final List<String> validationErrors =
+ JobResourceRequirements.validate(jobResourceRequirements, maxParallelismPerVertex);
+ assertThat(validationErrors).isEmpty();
+ }
+
+ @Test
+ void testWriteToJobGraphAndReadFromJobGraph() throws IOException {
+ final JobGraph jobGraph = JobGraphTestUtils.emptyJobGraph();
+ final JobResourceRequirements jobResourceRequirements =
+ JobResourceRequirements.newBuilder()
+ .setParallelismForJobVertex(firstVertexId, 10, 9)
+ .setParallelismForJobVertex(secondVertexId, 10, 10)
+ .build();
+
+ JobResourceRequirements.writeToJobGraph(jobGraph, jobResourceRequirements);
+ assertThat(JobResourceRequirements.readFromJobGraph(jobGraph))
+ .get()
+ .isEqualTo(jobResourceRequirements);
+ }
+
+ @Test
+ void testReadNonExistentResourceRequirementsFromJobGraph() throws IOException {
+ assertThat(JobResourceRequirements.readFromJobGraph(JobGraphTestUtils.emptyJobGraph()))
+ .isEmpty();
+ }
+}