[SPARK-18278][NOSUBMIT] Ongoing diff for Spark on Kubernetes (branch-2.2)

.github/ISSUE_TEMPLATE.md

@@ -0,0 +1,13 @@
+## Before you submit an issue
+The Kubernetes scheduler backend is currently being upstreamed to the main [Apache Spark project](https://github.com/apache/spark).
+We are attempting to re-direct as much new development as possible to the upstream.
+Please consider whether your issue can be submitted against the Apache Spark project,
+and submit there as an Apache Spark JRIA, if possible.
+
+If you have any questions about whether an issue should be submitted upstream or against this fork,
+please feel free to reach out on the following channels:
+
+* Apache Spark developer mailing list: [email protected]
+* Apache Spark [JIRA](https://issues.apache.org/jira/)
+* Big Data SIG [slack channel](https://kubernetes.slack.com/)
+* Regular Big Data SIG [meetings](https://github.com/kubernetes/community/tree/master/sig-big-data)

.github/PULL_REQUEST_TEMPLATE

@@ -1,3 +1,16 @@
+## Before you submit a Pull Request
+The Kubernetes scheduler backend is currently being upstreamed to the main [Apache Spark project](https://github.com/apache/spark).
+We are attempting to re-direct as much new development as possible to the upstream.
+Please consider whether your pull request can be submitted against the Apache Spark project, and submit there if possible.
+
+If you have any questions about whether a PR should be submitted upstream or against this fork,
+please feel free to reach out on the following channels:
+
+* Apache Spark developer mailing list: [email protected]
+* Apache Spark [JIRA](https://issues.apache.org/jira/)
+* Big Data SIG [slack channel](https://kubernetes.slack.com/)
+* Regular Big Data SIG [meetings](https://github.com/kubernetes/community/tree/master/sig-big-data)
+
 ## What changes were proposed in this pull request?
 
 (Please fill in changes proposed in this fix)

.travis.yml

@@ -25,10 +25,22 @@
 sudo: required
 dist: trusty
 
-# 2. Choose language and target JDKs for parallel builds.
+# 2. Choose language, target JDK and env's for parallel builds.
 language: java
 jdk:
   - oraclejdk8
+env:  # Used by the install section below.
+  # Configure the unit test build for spark core and kubernetes modules,
+  # while excluding some flaky unit tests using a regex pattern.
+  - PHASE=test  \
+    PROFILES="-Pmesos -Pyarn -Phadoop-2.7 -Pkubernetes"  \
+    MODULES="-pl core,resource-managers/kubernetes/core -am"  \
+    ARGS="-Dtest=none -Dsuffixes='^org\.apache\.spark\.(?!ExternalShuffleServiceSuite|SortShuffleSuite$|rdd\.LocalCheckpointSuite$|deploy\.SparkSubmitSuite$|deploy\.StandaloneDynamicAllocationSuite$).*'"
+  # Configure the full build.
+  - PHASE=install  \
+    PROFILES="-Pmesos -Pyarn -Phadoop-2.7 -Pkubernetes -Pkinesis-asl -Phive -Phive-thriftserver"  \
+    MODULES=""  \
+    ARGS="-T 4 -q -DskipTests"
 
 # 3. Setup cache directory for SBT and Maven.
 cache:
@@ -40,11 +52,12 @@ cache:
 notifications:
   email: false
 
-# 5. Run maven install before running lint-java.
+# 5. Run maven build before running lints.
 install:
   - export MAVEN_SKIP_RC=1
-  - build/mvn -T 4 -q -DskipTests -Pmesos -Pyarn -Pkinesis-asl -Phive -Phive-thriftserver install
+  - build/mvn ${PHASE} ${PROFILES} ${MODULES} ${ARGS}
 
-# 6. Run lint-java.
+# 6. Run lints.
 script:
   - dev/lint-java
+  - dev/lint-scala

README.md

@@ -1,3 +1,51 @@
+**ARCHIVED** This repository is being archived, to prevent any future confusion: All development on the Kubernetes scheduler back-end for Apache Spark is now upstream at https://spark.apache.org/ and https://github.com/apache/spark/
+
+**DEPRECATED**. Work on this fork is discontinued. Further development is continuing on the mainline implementation of Apache Spark: https://github.com/apache/spark.
+
+You can run Spark on Kubernetes using Spark 2.3. Some features from this work need to be ported to mainline. If a feature is missing, please check https://issues.apache.org/jira/projects/SPARK/issues to see if we're tracking that work, and if we are not, please file a JIRA ticket indicating the missing behavior.
+
+All other bugs and feature requests should either be proposed through JIRA or sent to [email protected] or [email protected].
+
+# Apache Spark On Kubernetes
+
+This repository, located at https://github.com/apache-spark-on-k8s/spark, contains a fork of Apache Spark that enables running Spark jobs natively on a Kubernetes cluster.
+
+## What is this?
+
+This is a collaboratively maintained project working on [SPARK-18278](https://issues.apache.org/jira/browse/SPARK-18278). The goal is to bring native support for Spark to use Kubernetes as a cluster manager, in a fully supported way on par with the Spark Standalone, Mesos, and Apache YARN cluster managers.
+
+## Getting Started
+
+- [Usage guide](https://apache-spark-on-k8s.github.io/userdocs/) shows how to run the code
+- [Development docs](resource-managers/kubernetes/README.md) shows how to get set up for development
+- [Architecture docs](resource-managers/kubernetes/architecture-docs/) shows the high level architecture of Spark on Kubernetes
+- Code is primarily located in the [resource-managers/kubernetes](resource-managers/kubernetes) folder
+
+## Why does this fork exist?
+
+Adding native integration for a new cluster manager is a large undertaking.  If poorly executed, it could introduce bugs into Spark when run on other cluster managers, cause release blockers slowing down the overall Spark project, or require hotfixes which divert attention away from development towards managing additional releases.  Any work this deep inside Spark needs to be done carefully to minimize the risk of those negative externalities.
+
+At the same time, an increasing number of people from various companies and organizations desire to work together to natively run Spark on Kubernetes.  The group needs a code repository, communication forum, issue tracking, and continuous integration, all in order to work together effectively on an open source product.
+
+We've been asked by an Apache Spark Committer to work outside of the Apache infrastructure for a short period of time to allow this feature to be hardened and improved without creating risk for Apache Spark.  The aim is to rapidly bring it to the point where it can be brought into the mainline Apache Spark repository for continued development within the Apache umbrella.  If all goes well, this should be a short-lived fork rather than a long-lived one.
+
+## Who are we?
+
+This is a collaborative effort by several folks from different companies who are interested in seeing this feature be successful.  Companies active in this project include (alphabetically):
+
+- Bloomberg
+- Google
+- Haiwen
+- Hyperpilot
+- Intel
+- Palantir
+- Pepperdata
+- Red Hat
+
+--------------------
+
+(original README below)
+
 # Apache Spark
 
 Spark is a fast and general cluster computing system for Big Data. It provides

assembly/pom.xml

@@ -21,7 +21,7 @@
   <parent>
     <groupId>org.apache.spark</groupId>
     <artifactId>spark-parent_2.11</artifactId>
-    <version>2.2.0</version>
+    <version>2.2.0-k8s-0.5.0</version>
     <relativePath>../pom.xml</relativePath>
   </parent>
 
@@ -148,6 +148,16 @@
         </dependency>
       </dependencies>
     </profile>
+    <profile>
+      <id>kubernetes</id>
+      <dependencies>
+        <dependency>
+          <groupId>org.apache.spark</groupId>
+          <artifactId>spark-kubernetes_${scala.binary.version}</artifactId>
+          <version>${project.version}</version>
+        </dependency>
+      </dependencies>
+    </profile>
     <profile>
       <id>hive</id>
       <dependencies>

common/network-common/pom.xml

@@ -22,7 +22,7 @@
   <parent>
     <groupId>org.apache.spark</groupId>
     <artifactId>spark-parent_2.11</artifactId>
-    <version>2.2.0</version>
+    <version>2.2.0-k8s-0.5.0</version>
     <relativePath>../../pom.xml</relativePath>
   </parent>
 

common/network-shuffle/pom.xml

@@ -22,7 +22,7 @@
   <parent>
     <groupId>org.apache.spark</groupId>
     <artifactId>spark-parent_2.11</artifactId>
-    <version>2.2.0</version>
+    <version>2.2.0-k8s-0.5.0</version>
     <relativePath>../../pom.xml</relativePath>
   </parent>
 

common/network-shuffle/src/main/java/org/apache/spark/network/shuffle/kubernetes/KubernetesExternalShuffleClient.java

@@ -0,0 +1,29 @@
+/*
+ * 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.spark.network.shuffle.kubernetes;
+
+import java.io.Closeable;
+import java.io.IOException;
+
+public interface KubernetesExternalShuffleClient extends Closeable {
+
+  void init(String appId);
+
+  void registerDriverWithShuffleService(String host, int port)
+      throws IOException, InterruptedException;
+}

common/network-shuffle/src/main/java/org/apache/spark/network/shuffle/kubernetes/KubernetesExternalShuffleClientImpl.java

@@ -0,0 +1,77 @@
+/*
+ * 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.spark.network.shuffle.kubernetes;
+
+import org.apache.spark.network.client.RpcResponseCallback;
+import org.apache.spark.network.client.TransportClient;
+import org.apache.spark.network.sasl.SecretKeyHolder;
+import org.apache.spark.network.shuffle.ExternalShuffleClient;
+import org.apache.spark.network.shuffle.protocol.RegisterDriver;
+import org.apache.spark.network.util.TransportConf;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import java.io.IOException;
+import java.nio.ByteBuffer;
+
+/**
+ * A client for talking to the external shuffle service in Kubernetes cluster mode.
+ *
+ * This is used by the each Spark executor to register with a corresponding external
+ * shuffle service on the cluster. The purpose is for cleaning up shuffle files
+ * reliably if the application exits unexpectedly.
+ */
+public class KubernetesExternalShuffleClientImpl
+    extends ExternalShuffleClient implements KubernetesExternalShuffleClient {
+
+  private static final Logger logger = LoggerFactory
+          .getLogger(KubernetesExternalShuffleClientImpl.class);
+
+  /**
+   * Creates a Kubernetes external shuffle client that wraps the {@link ExternalShuffleClient}.
+   * Please refer to docs on {@link ExternalShuffleClient} for more information.
+   */
+  public KubernetesExternalShuffleClientImpl(
+      TransportConf conf,
+      SecretKeyHolder secretKeyHolder,
+      boolean saslEnabled) {
+    super(conf, secretKeyHolder, saslEnabled);
+  }
+
+  @Override
+  public void registerDriverWithShuffleService(String host, int port)
+      throws IOException, InterruptedException {
+    checkInit();
+    ByteBuffer registerDriver = new RegisterDriver(appId, 0).toByteBuffer();
+    TransportClient client = clientFactory.createClient(host, port);
+    client.sendRpc(registerDriver, new RegisterDriverCallback());
+  }
+
+  private class RegisterDriverCallback implements RpcResponseCallback {
+    @Override
+    public void onSuccess(ByteBuffer response) {
+      logger.info("Successfully registered app " + appId + " with external shuffle service.");
+    }
+
+    @Override
+    public void onFailure(Throwable e) {
+      logger.warn("Unable to register app " + appId + " with external shuffle service. " +
+          "Please manually remove shuffle data after driver exit. Error: " + e);
+    }
+  }
+}

common/network-shuffle/src/main/java/org/apache/spark/network/shuffle/mesos/MesosExternalShuffleClient.java

@@ -32,7 +32,7 @@
 import org.apache.spark.network.client.TransportClient;
 import org.apache.spark.network.sasl.SecretKeyHolder;
 import org.apache.spark.network.shuffle.ExternalShuffleClient;
-import org.apache.spark.network.shuffle.protocol.mesos.RegisterDriver;
+import org.apache.spark.network.shuffle.protocol.RegisterDriver;
 import org.apache.spark.network.util.TransportConf;
 
 /**

common/network-shuffle/src/main/java/org/apache/spark/network/shuffle/protocol/BlockTransferMessage.java

@@ -23,7 +23,6 @@
 import io.netty.buffer.Unpooled;
 
 import org.apache.spark.network.protocol.Encodable;
-import org.apache.spark.network.shuffle.protocol.mesos.RegisterDriver;
 import org.apache.spark.network.shuffle.protocol.mesos.ShuffleServiceHeartbeat;
 
 /**

common/network-shuffle/src/main/java/org/apache/spark/network/shuffle/protocol/mesos/RegisterDriver.java → common/network-shuffle/src/main/java/org/apache/spark/network/shuffle/protocol/RegisterDriver.java

@@ -15,19 +15,18 @@
  * limitations under the License.
  */
 
-package org.apache.spark.network.shuffle.protocol.mesos;
+package org.apache.spark.network.shuffle.protocol;
 
 import com.google.common.base.Objects;
 import io.netty.buffer.ByteBuf;
 
 import org.apache.spark.network.protocol.Encoders;
-import org.apache.spark.network.shuffle.protocol.BlockTransferMessage;
 
 // Needed by ScalaDoc. See SPARK-7726
 import static org.apache.spark.network.shuffle.protocol.BlockTransferMessage.Type;
 
 /**
- * A message sent from the driver to register with the MesosExternalShuffleService.
+ * A message sent from the driver to register with an ExternalShuffleService.
  */
 public class RegisterDriver extends BlockTransferMessage {
   private final String appId;

common/network-yarn/pom.xml

@@ -22,7 +22,7 @@
   <parent>
     <groupId>org.apache.spark</groupId>
     <artifactId>spark-parent_2.11</artifactId>
-    <version>2.2.0</version>
+    <version>2.2.0-k8s-0.5.0</version>
     <relativePath>../../pom.xml</relativePath>
   </parent>
 

common/sketch/pom.xml

@@ -22,7 +22,7 @@
   <parent>
     <groupId>org.apache.spark</groupId>
     <artifactId>spark-parent_2.11</artifactId>
-    <version>2.2.0</version>
+    <version>2.2.0-k8s-0.5.0</version>
     <relativePath>../../pom.xml</relativePath>
   </parent>
 

common/tags/pom.xml

@@ -22,7 +22,7 @@
   <parent>
     <groupId>org.apache.spark</groupId>
     <artifactId>spark-parent_2.11</artifactId>
-    <version>2.2.0</version>
+    <version>2.2.0-k8s-0.5.0</version>
     <relativePath>../../pom.xml</relativePath>
   </parent>
 

common/unsafe/pom.xml

@@ -22,7 +22,7 @@
   <parent>
     <groupId>org.apache.spark</groupId>
     <artifactId>spark-parent_2.11</artifactId>
-    <version>2.2.0</version>
+    <version>2.2.0-k8s-0.5.0</version>
     <relativePath>../../pom.xml</relativePath>
   </parent>