sokui commented on code in PR #3186: URL: https://github.com/apache/ozone/pull/3186#discussion_r888438128
########## hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/node/NodeAddressUpdateHandler.java: ########## @@ -0,0 +1,69 @@ +/** + * 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 + * <p> + * http://www.apache.org/licenses/LICENSE-2.0 + * <p> + * 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.hadoop.hdds.scm.node; + +import org.apache.hadoop.hdds.protocol.DatanodeDetails; +import org.apache.hadoop.hdds.scm.ha.SCMService; +import org.apache.hadoop.hdds.scm.ha.SCMServiceManager; +import org.apache.hadoop.hdds.scm.node.states.NodeNotFoundException; +import org.apache.hadoop.hdds.scm.pipeline.PipelineManager; +import org.apache.hadoop.hdds.server.events.EventHandler; +import org.apache.hadoop.hdds.server.events.EventPublisher; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +/** + * Handles datanode ip or hostname change event. + */ +public class NodeAddressUpdateHandler + implements EventHandler<DatanodeDetails> { + private static final Logger LOG = + LoggerFactory.getLogger(NodeAddressUpdateHandler.class); + + private final PipelineManager pipelineManager; + private final NodeDecommissionManager decommissionManager; + private final SCMServiceManager serviceManager; + + public NodeAddressUpdateHandler(PipelineManager pipelineManager, + NodeDecommissionManager + decommissionManager, + SCMServiceManager serviceManager) { + this.pipelineManager = pipelineManager; + this.decommissionManager = decommissionManager; + this.serviceManager = serviceManager; + } + + @Override + public void onMessage(DatanodeDetails datanodeDetails, + EventPublisher publisher) { + try { + LOG.info("Closing stale pipelines for datanode: {}", datanodeDetails); + pipelineManager.closeStalePipelines(datanodeDetails); Review Comment: Ah, I got what you mean. This is because the restarts are so quick that SCM thinks these 3 datanodes keep alive. So when the first DN re-registers and tries to create the new pipeline, SCM thinks the other two DN are still good, and thus uses the old info (because the other two DNs have not re-registered yet). Same thing happens when the 2nd DN re-registeres, where SCM thinks the 3rd DN is good and uses its old info. For quick restarts, I do not think there is a good way for SCM to recognize these DN's are dead/restarted. Therefore, we rely on the logic of this PR for waiting ALLOCATED pipelines to be OPEN. In such way, the block allocation can still get the correct pipeline eventually. -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
