[FLINK-6136] Separate EmbeddedHaServices and StandaloneHaServices

This PR introduces a standalone high availability services implementation which can be used in a distributed setting with no HA guarantees. Additionally, it introduces a common base class which is also used by the EmbeddedHaServices. This base class instantiates the standalone variants of the checkpoint recovery factory, submitted job graphs store, running jobs registry and blob store. The StandaloneHaServices are instantiated with a fixed address for the Job- and ResourceManager. This address and the HighAvailability.DEFAULT_LEADER_ID is returned by the corresponding LeaderRetrievalServices when being started. This closes apache#3622.
PangZhi · May 5, 2017 · a0bb99c · a0bb99c
1 parent 43fa507
commit a0bb99c
Show file tree

Hide file tree

Showing 62 changed files with 897 additions and 581 deletions.
diff --git a/flink-clients/src/main/java/org/apache/flink/client/program/ClusterClient.java b/flink-clients/src/main/java/org/apache/flink/client/program/ClusterClient.java
@@ -241,7 +241,7 @@ public InetSocketAddress getJobManagerAddress() {
 		try {
 			LeaderConnectionInfo leaderConnectionInfo =
 				LeaderRetrievalUtils.retrieveLeaderConnectionInfo(
-					LeaderRetrievalUtils.createLeaderRetrievalService(flinkConfig), timeout);
+					LeaderRetrievalUtils.createLeaderRetrievalService(flinkConfig, true), timeout);
 
 			return AkkaUtils.getInetSockeAddressFromAkkaURL(leaderConnectionInfo.getAddress());
 		} catch (Exception e) {
@@ -464,7 +464,7 @@ public JobSubmissionResult runDetached(JobGraph jobGraph, ClassLoader classLoade
 	public JobExecutionResult retrieveJob(JobID jobID) throws JobExecutionException {
 		final LeaderRetrievalService leaderRetrievalService;
 		try {
-			leaderRetrievalService = LeaderRetrievalUtils.createLeaderRetrievalService(flinkConfig);
+			leaderRetrievalService = LeaderRetrievalUtils.createLeaderRetrievalService(flinkConfig, true);
 		} catch (Exception e) {
 			throw new JobRetrievalException(jobID, "Could not create the leader retrieval service", e);
 		}
@@ -498,7 +498,7 @@ public JobExecutionResult retrieveJob(JobID jobID) throws JobExecutionException
 	public JobListeningContext connectToJob(JobID jobID) throws JobExecutionException {
 		final LeaderRetrievalService leaderRetrievalService;
 		try {
-			leaderRetrievalService = LeaderRetrievalUtils.createLeaderRetrievalService(flinkConfig);
+			leaderRetrievalService = LeaderRetrievalUtils.createLeaderRetrievalService(flinkConfig, true);
 		} catch (Exception e) {
 			throw new JobRetrievalException(jobID, "Could not create the leader retrieval service", e);
 		}
@@ -721,7 +721,7 @@ private JobGraph getJobGraph(FlinkPlan optPlan, List<URL> jarFiles, List<URL> cl
 	public ActorGateway getJobManagerGateway() throws Exception {
 		LOG.debug("Looking up JobManager");
 		return LeaderRetrievalUtils.retrieveLeaderGateway(
-			LeaderRetrievalUtils.createLeaderRetrievalService(flinkConfig),
+			LeaderRetrievalUtils.createLeaderRetrievalService(flinkConfig, true),
 			actorSystemLoader.get(),
 			lookupTimeout);
 	}

diff --git a/...k-clients/src/test/java/org/apache/flink/client/RemoteExecutorHostnameResolutionTest.java b/...k-clients/src/test/java/org/apache/flink/client/RemoteExecutorHostnameResolutionTest.java
@@ -23,6 +23,7 @@
 import org.apache.flink.api.java.io.DiscardingOutputFormat;
 import org.apache.flink.client.program.ProgramInvocationException;
 import org.apache.flink.configuration.Configuration;
+import org.apache.flink.util.TestLogger;
 import org.junit.BeforeClass;
 import org.junit.Test;
 
@@ -36,7 +37,7 @@
 import static org.junit.Assert.fail;
 import static org.junit.Assume.assumeTrue;
 
-public class RemoteExecutorHostnameResolutionTest {
+public class RemoteExecutorHostnameResolutionTest extends TestLogger {
 
 	private static final String nonExistingHostname = "foo.bar.com.invalid";
 	private static final int port = 14451;

diff --git a/flink-clients/src/test/java/org/apache/flink/client/program/ClientTest.java b/flink-clients/src/test/java/org/apache/flink/client/program/ClientTest.java
@@ -43,7 +43,7 @@
 import org.apache.flink.runtime.akka.FlinkUntypedActor;
 import org.apache.flink.api.common.JobID;
 import org.apache.flink.runtime.highavailability.HighAvailabilityServices;
-import org.apache.flink.runtime.jobmanager.JobManager;
+import org.apache.flink.runtime.jobmaster.JobMaster;
 import org.apache.flink.runtime.messages.JobManagerMessages;
 import org.apache.flink.runtime.util.SerializedThrowable;
 import org.apache.flink.util.NetUtils;
@@ -67,7 +67,6 @@
 import static org.mockito.Mockito.mock;
 import static org.mockito.Mockito.when;
 
-
 /**
  * Simple and maybe stupid test to check the {@link ClusterClient} class.
  */
@@ -129,7 +128,9 @@ public void shutDownActorSystem() {
 	 */
 	@Test
 	public void testDetachedMode() throws Exception{
-		jobManagerSystem.actorOf(Props.create(SuccessReturningActor.class), JobManager.JOB_MANAGER_NAME());
+		jobManagerSystem.actorOf(
+			Props.create(SuccessReturningActor.class),
+			JobMaster.JOB_MANAGER_NAME);
 		ClusterClient out = new StandaloneClusterClient(config);
 		out.setDetached(true);
 
@@ -198,30 +199,28 @@ public void testDetachedMode() throws Exception{
 	 * This test verifies correct job submission messaging logic and plan translation calls.
 	 */
 	@Test
-	public void shouldSubmitToJobClient() {
-		try {
-			jobManagerSystem.actorOf(Props.create(SuccessReturningActor.class), JobManager.JOB_MANAGER_NAME());
+	public void shouldSubmitToJobClient() throws IOException, ProgramInvocationException {
+		jobManagerSystem.actorOf(
+			Props.create(SuccessReturningActor.class),
+			JobMaster.JOB_MANAGER_NAME);
 
-			ClusterClient out = new StandaloneClusterClient(config);
-			out.setDetached(true);
-			JobSubmissionResult result = out.run(program.getPlanWithJars(), 1);
+		ClusterClient out = new StandaloneClusterClient(config);
+		out.setDetached(true);
+		JobSubmissionResult result = out.run(program.getPlanWithJars(), 1);
 
-			assertNotNull(result);
+		assertNotNull(result);
 
-			program.deleteExtractedLibraries();
-		}
-		catch (Exception e) {
-			e.printStackTrace();
-			fail(e.getMessage());
-		}
+		program.deleteExtractedLibraries();
 	}
 
 	/**
 	 * This test verifies correct that the correct exception is thrown when the job submission fails.
 	 */
 	@Test
 	public void shouldSubmitToJobClientFails() throws IOException {
-		jobManagerSystem.actorOf(Props.create(FailureReturningActor.class), JobManager.JOB_MANAGER_NAME());
+			jobManagerSystem.actorOf(
+				Props.create(FailureReturningActor.class),
+				JobMaster.JOB_MANAGER_NAME);
 
 		ClusterClient out = new StandaloneClusterClient(config);
 		out.setDetached(true);
@@ -245,7 +244,9 @@ public void shouldSubmitToJobClientFails() throws IOException {
 	@Test
 	public void tryLocalExecution() {
 		try {
-			jobManagerSystem.actorOf(Props.create(SuccessReturningActor.class), JobManager.JOB_MANAGER_NAME());
+			jobManagerSystem.actorOf(
+				Props.create(SuccessReturningActor.class),
+				JobMaster.JOB_MANAGER_NAME);
 
 			PackagedProgram packagedProgramMock = mock(PackagedProgram.class);
 			when(packagedProgramMock.isUsingInteractiveMode()).thenReturn(true);
@@ -276,7 +277,9 @@ public Void answer(InvocationOnMock invocation) throws Throwable {
 	@Test
 	public void testGetExecutionPlan() {
 		try {
-			jobManagerSystem.actorOf(Props.create(FailureReturningActor.class), JobManager.JOB_MANAGER_NAME());
+			jobManagerSystem.actorOf(
+				Props.create(FailureReturningActor.class),
+				JobMaster.JOB_MANAGER_NAME);
 
 			PackagedProgram prg = new PackagedProgram(TestOptimizerPlan.class, "/dev/random", "/tmp");
 			assertNotNull(prg.getPreviewPlan());

diff --git a/...st/java/org/apache/flink/client/program/LeaderRetrievalServiceHostnameResolutionTest.java b/...st/java/org/apache/flink/client/program/LeaderRetrievalServiceHostnameResolutionTest.java
@@ -56,7 +56,7 @@ public void testUnresolvableHostname1() {
 			config.setString(ConfigConstants.JOB_MANAGER_IPC_ADDRESS_KEY, nonExistingHostname);
 			config.setInteger(ConfigConstants.JOB_MANAGER_IPC_PORT_KEY, 17234);
 
-			LeaderRetrievalUtils.createLeaderRetrievalService(config);
+			LeaderRetrievalUtils.createLeaderRetrievalService(config, false);
 		}
 		catch (Exception e) {
 			System.err.println("Shouldn't throw an exception!");
@@ -69,7 +69,7 @@ public void testUnresolvableHostname1() {
 	 * Tests that the StandaloneLeaderRetrievalService does not resolve host names by default.
 	 */
 	@Test
-	public void testUnresolvableHostname2() {
+	public void testUnresolvableHostname2() throws Exception {
 
 		try {
 			Configuration config = new Configuration();
@@ -78,16 +78,11 @@ public void testUnresolvableHostname2() {
 			config.setInteger(ConfigConstants.JOB_MANAGER_IPC_PORT_KEY, 17234);
 
 			LeaderRetrievalUtils.createLeaderRetrievalService(config, true);
-			fail("This should fail with an IllegalConfigurationException");
+			fail("This should fail with an UnknownHostException");
 		}
 		catch (UnknownHostException e) {
 			// that is what we want!
 		}
-		catch (Exception e) {
-			System.err.println("Wrong exception!");
-			e.printStackTrace();
-			fail(e.getMessage());
-		}
 	}
 
 	private static void checkPreconditions() {

diff --git a/flink-contrib/flink-storm/src/main/java/org/apache/flink/storm/api/FlinkClient.java b/flink-contrib/flink-storm/src/main/java/org/apache/flink/storm/api/FlinkClient.java
@@ -22,6 +22,9 @@
 import akka.actor.ActorSystem;
 import akka.pattern.Patterns;
 import akka.util.Timeout;
+import org.apache.flink.runtime.highavailability.HighAvailabilityServicesUtils.AddressResolution;
+import org.apache.flink.runtime.jobmaster.JobMaster;
+import org.apache.flink.runtime.rpc.akka.AkkaRpcServiceUtils;
 import org.apache.storm.Config;
 import org.apache.storm.generated.AlreadyAliveException;
 import org.apache.storm.generated.InvalidTopologyException;
@@ -43,12 +46,10 @@
 import org.apache.flink.runtime.akka.AkkaUtils;
 import org.apache.flink.runtime.client.JobStatusMessage;
 import org.apache.flink.runtime.jobgraph.JobGraph;
-import org.apache.flink.runtime.jobmanager.JobManager;
 import org.apache.flink.runtime.messages.JobManagerMessages;
 import org.apache.flink.runtime.messages.JobManagerMessages.RunningJobsStatus;
 import org.apache.flink.storm.util.StormConfig;
 import org.apache.flink.streaming.api.graph.StreamGraph;
-import org.apache.flink.util.NetUtils;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 import scala.Some;
@@ -326,9 +327,14 @@ private ActorRef getJobManager() throws IOException {
 			throw new RuntimeException("Could not start actor system to communicate with JobManager", e);
 		}
 
-		return JobManager.getJobManagerActorRef(AkkaUtils.getAkkaProtocol(configuration),
-				NetUtils.unresolvedHostAndPortToNormalizedString(this.jobManagerHost, this.jobManagerPort),
-				actorSystem, AkkaUtils.getLookupTimeout(configuration));
+		final String jobManagerAkkaUrl = AkkaRpcServiceUtils.getRpcUrl(
+			jobManagerHost,
+			jobManagerPort,
+			JobMaster.JOB_MANAGER_NAME,
+			AddressResolution.TRY_ADDRESS_RESOLUTION,
+			configuration);
+
+		return AkkaUtils.getActorRef(jobManagerAkkaUrl, actorSystem, AkkaUtils.getLookupTimeout(configuration));
 	}
 
 	@SuppressWarnings({ "unchecked", "rawtypes" })

diff --git a/...er/exceptions/ConfigurationException.java → ...he/flink/util/ConfigurationException.java b/...er/exceptions/ConfigurationException.java → ...he/flink/util/ConfigurationException.java
@@ -16,12 +16,12 @@
  * limitations under the License.
  */
 
-package org.apache.flink.runtime.resourcemanager.exceptions;
+package org.apache.flink.util;
 
 /**
- * Exception which occures when creating a configuration object fails.
+ * Exception which occurs when creating a configuration object fails.
  */
-public class ConfigurationException extends Exception {
+public class ConfigurationException extends FlinkException {
 	private static final long serialVersionUID = 3971647332059381556L;
 
 	public ConfigurationException(String message) {

diff --git a/...in/java/org/apache/flink/mesos/runtime/clusterframework/MesosApplicationMasterRunner.java b/...in/java/org/apache/flink/mesos/runtime/clusterframework/MesosApplicationMasterRunner.java
@@ -50,6 +50,7 @@
 import org.apache.flink.runtime.clusterframework.overlays.SSLStoreOverlay;
 import org.apache.flink.runtime.jobmanager.JobManager;
 import org.apache.flink.runtime.jobmanager.MemoryArchivist;
+import org.apache.flink.runtime.jobmaster.JobMaster;
 import org.apache.flink.runtime.leaderretrieval.LeaderRetrievalService;
 import org.apache.flink.runtime.process.ProcessReaper;
 import org.apache.flink.runtime.security.SecurityUtils;
@@ -65,6 +66,8 @@
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
+import scala.Option;
+import scala.Some;
 import scala.concurrent.duration.Duration;
 import scala.concurrent.duration.FiniteDuration;
 
@@ -301,8 +304,8 @@ protected int runPrivileged(Configuration config, Configuration dynamicPropertie
 				actorSystem,
 				futureExecutor,
 				ioExecutor,
-				new scala.Some<>(JobManager.JOB_MANAGER_NAME()),
-				scala.Option.<String>empty(),
+				new Some<>(JobMaster.JOB_MANAGER_NAME),
+				Option.<String>empty(),
 				getJobManagerClass(),
 				getArchivistClass())._1();
 

diff --git a/...-web/src/main/java/org/apache/flink/runtime/webmonitor/handlers/HandlerRedirectUtils.java b/...-web/src/main/java/org/apache/flink/runtime/webmonitor/handlers/HandlerRedirectUtils.java
@@ -25,12 +25,11 @@
 import io.netty.handler.codec.http.HttpResponseStatus;
 import io.netty.handler.codec.http.HttpVersion;
 import org.apache.flink.configuration.ConfigConstants;
+import org.apache.flink.runtime.akka.AkkaUtils;
 import org.apache.flink.runtime.instance.ActorGateway;
-import org.apache.flink.runtime.jobmanager.JobManager;
 import org.apache.flink.runtime.webmonitor.files.MimeTypes;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
-import scala.Option;
 import scala.Tuple2;
 
 import java.util.regex.Matcher;
@@ -62,7 +61,7 @@ public static String getRedirectAddress(
 		final String jobManagerName = localJobManagerAddress.substring(localJobManagerAddress.lastIndexOf("/") + 1);
 
 		if (!localJobManagerAddress.equals(leaderAddress) &&
-			!leaderAddress.equals(JobManager.getLocalJobManagerAkkaURL(Option.apply(jobManagerName)))) {
+			!leaderAddress.equals(AkkaUtils.getLocalAkkaURL(jobManagerName))) {
 			// We are not the leader and need to redirect
 			Matcher matcher = LeaderAddressHostPattern.matcher(leaderAddress);
 

diff --git a/flink-runtime/src/main/java/org/apache/flink/runtime/blob/BlobUtils.java b/flink-runtime/src/main/java/org/apache/flink/runtime/blob/BlobUtils.java
@@ -24,7 +24,7 @@
 import org.apache.flink.configuration.ConfigConstants;
 import org.apache.flink.configuration.Configuration;
 import org.apache.flink.configuration.IllegalConfigurationException;
-import org.apache.flink.runtime.highavailability.ZookeeperHaServices;
+import org.apache.flink.runtime.highavailability.zookeeper.ZooKeeperHaServices;
 import org.apache.flink.runtime.jobmanager.HighAvailabilityMode;
 import org.apache.flink.util.StringUtils;
 import org.slf4j.Logger;
@@ -84,7 +84,7 @@ static BlobStore createBlobStoreFromConfig(Configuration config) throws IOExcept
 		if (highAvailabilityMode == HighAvailabilityMode.NONE) {
 			return new VoidBlobStore();
 		} else if (highAvailabilityMode == HighAvailabilityMode.ZOOKEEPER) {
-			return ZookeeperHaServices.createBlobStore(config);
+			return ZooKeeperHaServices.createBlobStore(config);
 		} else {
 			throw new IllegalConfigurationException("Unexpected high availability mode '" + highAvailabilityMode + "'.");
 		}

diff --git a/flink-runtime/src/main/java/org/apache/flink/runtime/client/JobListeningContext.java b/flink-runtime/src/main/java/org/apache/flink/runtime/client/JobListeningContext.java
@@ -135,7 +135,7 @@ public ClassLoader getClassLoader() throws JobRetrievalException {
 	private ActorGateway getJobManager() throws JobRetrievalException {
 		try {
 			return LeaderRetrievalUtils.retrieveLeaderGateway(
-				LeaderRetrievalUtils.createLeaderRetrievalService(configuration),
+				LeaderRetrievalUtils.createLeaderRetrievalService(configuration, true),
 				actorSystem,
 				AkkaUtils.getLookupTimeout(configuration));
 		} catch (Exception e) {

diff --git a/...untime/src/main/java/org/apache/flink/runtime/highavailability/EmbeddedNonHaServices.java b/...untime/src/main/java/org/apache/flink/runtime/highavailability/EmbeddedNonHaServices.java