wangyang0918 commented on a change in pull request #13644: URL: https://github.com/apache/flink/pull/13644#discussion_r514063391
########## File path: flink-kubernetes/src/main/java/org/apache/flink/kubernetes/highavailability/KubernetesLeaderElectionDriver.java ########## @@ -0,0 +1,279 @@ +/* + * 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.kubernetes.highavailability; + +import org.apache.flink.kubernetes.configuration.KubernetesLeaderElectionConfiguration; +import org.apache.flink.kubernetes.kubeclient.FlinkKubeClient; +import org.apache.flink.kubernetes.kubeclient.resources.KubernetesConfigMap; +import org.apache.flink.kubernetes.kubeclient.resources.KubernetesException; +import org.apache.flink.kubernetes.kubeclient.resources.KubernetesLeaderElector; +import org.apache.flink.kubernetes.kubeclient.resources.KubernetesWatch; +import org.apache.flink.kubernetes.utils.KubernetesUtils; +import org.apache.flink.runtime.concurrent.FutureUtils; +import org.apache.flink.runtime.leaderelection.LeaderElectionDriver; +import org.apache.flink.runtime.leaderelection.LeaderElectionEventHandler; +import org.apache.flink.runtime.leaderelection.LeaderElectionException; +import org.apache.flink.runtime.leaderelection.LeaderInformation; +import org.apache.flink.runtime.rpc.FatalErrorHandler; + +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +import javax.annotation.Nullable; +import javax.annotation.concurrent.GuardedBy; + +import java.util.HashMap; +import java.util.List; +import java.util.Map; +import java.util.Optional; +import java.util.UUID; +import java.util.concurrent.CompletableFuture; +import java.util.concurrent.Executor; + +import static org.apache.flink.kubernetes.utils.Constants.LABEL_CONFIGMAP_TYPE_HIGH_AVAILABILITY; +import static org.apache.flink.kubernetes.utils.Constants.LEADER_ADDRESS_KEY; +import static org.apache.flink.kubernetes.utils.Constants.LEADER_SESSION_ID_KEY; +import static org.apache.flink.kubernetes.utils.KubernetesUtils.checkConfigMaps; +import static org.apache.flink.kubernetes.utils.KubernetesUtils.getLeaderInformationFromConfigMap; +import static org.apache.flink.util.Preconditions.checkNotNull; + +/** + * {@link LeaderElectionDriver} implementation for Kubernetes. The active leader is elected using Kubernetes. + * The current leader's address as well as its leader session ID is published via Kubernetes ConfigMap. + * Note that the contending lock and leader storage are using the same ConfigMap. And every component(e.g. + * ResourceManager, Dispatcher, RestEndpoint, JobManager for each job) will have a separate ConfigMap. + */ +public class KubernetesLeaderElectionDriver implements LeaderElectionDriver { + + private static final Logger LOG = LoggerFactory.getLogger(KubernetesLeaderElectionDriver.class); + + private final Object lock = new Object(); + + private final FlinkKubeClient kubeClient; + + private final Executor ioExecutor; + + private final String configMapName; + + private final String lockIdentity; + + private final KubernetesLeaderElector leaderElector; + + // Labels will be used to clean up the ha related ConfigMaps. + private final Map<String, String> configMapLabels; + + private final LeaderElectionEventHandler leaderElectionEventHandler; + + private final KubernetesWatch kubernetesWatch; + + private final FatalErrorHandler fatalErrorHandler; + + @GuardedBy("lock") + private volatile CompletableFuture<Void> leaderElectorRunFuture; + + @GuardedBy("lock") + private volatile boolean running; + + public KubernetesLeaderElectionDriver( + FlinkKubeClient kubeClient, + Executor ioExecutor, + KubernetesLeaderElectionConfiguration leaderConfig, + LeaderElectionEventHandler leaderElectionEventHandler, + FatalErrorHandler fatalErrorHandler) { + + this.kubeClient = checkNotNull(kubeClient, "Kubernetes client"); + this.ioExecutor = checkNotNull(ioExecutor, "IO Executor"); + + this.leaderElectionEventHandler = checkNotNull(leaderElectionEventHandler, "LeaderElectionEventHandler"); + this.fatalErrorHandler = checkNotNull(fatalErrorHandler); + + this.configMapName = leaderConfig.getConfigMapName(); + this.lockIdentity = leaderConfig.getLockIdentity(); + this.leaderElector = kubeClient.createLeaderElector(leaderConfig, new LeaderCallbackHandlerImpl()); + this.configMapLabels = KubernetesUtils.getConfigMapLabels( + leaderConfig.getClusterId(), LABEL_CONFIGMAP_TYPE_HIGH_AVAILABILITY); + + leaderElectorRunFuture = CompletableFuture.runAsync(leaderElector::run, ioExecutor); + FutureUtils.assertNoException(leaderElectorRunFuture); + kubernetesWatch = kubeClient.watchConfigMaps(configMapName, new ConfigMapCallbackHandlerImpl()); + running = true; + } + + @Override + public void close() { + synchronized (lock) { + if (!running) { + return; + } + running = false; + + LOG.info("Closing {}.", this); + leaderElectorRunFuture.complete(null); + kubernetesWatch.close(); + } + } + + @Override + public void writeLeaderInformation(LeaderInformation leaderInformation) { + final UUID confirmedLeaderSessionID = leaderInformation.getLeaderSessionID(); + final String confirmedLeaderAddress = leaderInformation.getLeaderAddress(); + try { + kubeClient.checkAndUpdateConfigMap( + configMapName, + configMap -> { + if (KubernetesLeaderElector.hasLeadership(configMap, lockIdentity)) { + // Get the updated ConfigMap with new leader information + if (confirmedLeaderAddress != null && confirmedLeaderSessionID != null) { + configMap.getData().put(LEADER_ADDRESS_KEY, confirmedLeaderAddress); + configMap.getData().put(LEADER_SESSION_ID_KEY, confirmedLeaderSessionID.toString()); + } + configMap.getLabels().putAll(configMapLabels); + return Optional.of(configMap); + } + return Optional.empty(); + }).get(); + } catch (Exception e) { + fatalErrorHandler.onFatalError( + new KubernetesException("Could not update ConfigMap " + configMapName, e)); + } + } + + @Override + public boolean hasLeadership() { + final Optional<KubernetesConfigMap> configMapOpt = kubeClient.getConfigMap(configMapName); + if (configMapOpt.isPresent()) { + return KubernetesLeaderElector.hasLeadership(configMapOpt.get(), lockIdentity); + } else { + fatalErrorHandler.onFatalError( + new KubernetesException("ConfigMap " + configMapName + "does not exist.", null)); + return false; + } + } + + private class LeaderCallbackHandlerImpl extends KubernetesLeaderElector.LeaderCallbackHandler { + + @Override + public void isLeader() { + synchronized (lock) { + if (running) { + leaderElectionEventHandler.onGrantLeadership(); + } + } + } + + @Override + public void notLeader() { + synchronized (lock) { + if (running) { + // Clear the leader information in ConfigMap + try { + kubeClient.checkAndUpdateConfigMap( + configMapName, + configMap -> { + if (KubernetesLeaderElector.hasLeadership(configMap, lockIdentity)) { + configMap.getData().remove(LEADER_ADDRESS_KEY); + configMap.getData().remove(LEADER_SESSION_ID_KEY); + return Optional.of(configMap); + } + return Optional.empty(); + } + ).get(); + } catch (Exception e) { + fatalErrorHandler.onFatalError( + new LeaderElectionException( + "Could not remove leader information from ConfigMap " + configMapName, e)); + } + leaderElectionEventHandler.onRevokeLeadership(); + // Continue to contend the leader + if (!leaderElectorRunFuture.isDone()) { + leaderElectorRunFuture.complete(null); + } + leaderElectorRunFuture = CompletableFuture.runAsync(leaderElector::run, ioExecutor); + FutureUtils.assertNoException(leaderElectorRunFuture); + } + } + } + } + + private class ConfigMapCallbackHandlerImpl implements FlinkKubeClient.WatchCallbackHandler<KubernetesConfigMap> { + // This is used to get the difference between current and previous data. Since the annotation will be updated + // periodically, we need to filter out these noises. + final Map<String, String> previousData = new HashMap<>(); + + @Override + public void onAdded(List<KubernetesConfigMap> configMaps) { + // noop + } + + @Override + public void onModified(List<KubernetesConfigMap> configMaps) { + // We should only receive events for the watched ConfigMap + final KubernetesConfigMap configMap = checkConfigMaps(configMaps, configMapName); + + if (!configMap.getData().equals(previousData)) { + if (KubernetesLeaderElector.hasLeadership(configMap, lockIdentity)) { + synchronized (lock) { + if (running) { + leaderElectionEventHandler.onLeaderInformationChange( + getLeaderInformationFromConfigMap(configMap)); Review comment: I think the implementor of `onLeaderInformationChange` should check whether the `LeaderInformation` has changed truly. We have already done this in the `DefaultLeaderElectionService`. I will remove the pre-check here and add a description of `LeaderElectionEventHandler#onLeaderInformationChange`. Duplicated leader change event could happen, the implementor should check this. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org
