001/** 002 * Licensed to the Apache Software Foundation (ASF) under one 003 * or more contributor license agreements. See the NOTICE file 004 * distributed with this work for additional information 005 * regarding copyright ownership. The ASF licenses this file 006 * to you under the Apache License, Version 2.0 (the 007 * "License"); you may not use this file except in compliance 008 * with the License. You may obtain a copy of the License at 009 * 010 * http://www.apache.org/licenses/LICENSE-2.0 011 * 012 * Unless required by applicable law or agreed to in writing, software 013 * distributed under the License is distributed on an "AS IS" BASIS, 014 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 015 * See the License for the specific language governing permissions and 016 * limitations under the License. 017 */ 018package org.apache.hadoop.hbase.regionserver.handler; 019 020import edu.umd.cs.findbugs.annotations.Nullable; 021import java.io.IOException; 022import java.util.concurrent.TimeUnit; 023import org.apache.hadoop.hbase.HConstants; 024import org.apache.hadoop.hbase.client.RegionInfo; 025import org.apache.hadoop.hbase.client.TableDescriptor; 026import org.apache.hadoop.hbase.executor.EventHandler; 027import org.apache.hadoop.hbase.executor.EventType; 028import org.apache.hadoop.hbase.regionserver.HRegion; 029import org.apache.hadoop.hbase.regionserver.HRegionServer; 030import org.apache.hadoop.hbase.regionserver.Region; 031import org.apache.hadoop.hbase.regionserver.RegionServerServices.PostOpenDeployContext; 032import org.apache.hadoop.hbase.regionserver.RegionServerServices.RegionStateTransitionContext; 033import org.apache.hadoop.hbase.util.RetryCounter; 034import org.apache.yetus.audience.InterfaceAudience; 035import org.slf4j.Logger; 036import org.slf4j.LoggerFactory; 037 038import org.apache.hadoop.hbase.shaded.protobuf.generated.RegionServerStatusProtos.RegionStateTransition.TransitionCode; 039 040/** 041 * Handles opening of a region on a region server. 042 * <p/> 043 * Just done the same thing with the old {@link OpenRegionHandler}, with some modifications on 044 * fencing and retrying. But we need to keep the {@link OpenRegionHandler} as is to keep compatible 045 * with the zk less assignment for 1.x, otherwise it is not possible to do rolling upgrade. 046 */ 047@InterfaceAudience.Private 048public class AssignRegionHandler extends EventHandler { 049 050 private static final Logger LOG = LoggerFactory.getLogger(AssignRegionHandler.class); 051 052 private final RegionInfo regionInfo; 053 054 private final long openProcId; 055 056 private final TableDescriptor tableDesc; 057 058 private final long masterSystemTime; 059 060 private final RetryCounter retryCounter; 061 062 public AssignRegionHandler(HRegionServer server, RegionInfo regionInfo, long openProcId, 063 @Nullable TableDescriptor tableDesc, long masterSystemTime, EventType eventType) { 064 super(server, eventType); 065 this.regionInfo = regionInfo; 066 this.openProcId = openProcId; 067 this.tableDesc = tableDesc; 068 this.masterSystemTime = masterSystemTime; 069 this.retryCounter = HandlerUtil.getRetryCounter(); 070 } 071 072 private HRegionServer getServer() { 073 return (HRegionServer) server; 074 } 075 076 private void cleanUpAndReportFailure(IOException error) throws IOException { 077 LOG.warn("Failed to open region {}, will report to master", regionInfo.getRegionNameAsString(), 078 error); 079 HRegionServer rs = getServer(); 080 rs.getRegionsInTransitionInRS().remove(regionInfo.getEncodedNameAsBytes(), Boolean.TRUE); 081 if (!rs.reportRegionStateTransition(new RegionStateTransitionContext(TransitionCode.FAILED_OPEN, 082 HConstants.NO_SEQNUM, openProcId, masterSystemTime, regionInfo))) { 083 throw new IOException( 084 "Failed to report failed open to master: " + regionInfo.getRegionNameAsString()); 085 } 086 } 087 088 @Override 089 public void process() throws IOException { 090 HRegionServer rs = getServer(); 091 String encodedName = regionInfo.getEncodedName(); 092 byte[] encodedNameBytes = regionInfo.getEncodedNameAsBytes(); 093 String regionName = regionInfo.getRegionNameAsString(); 094 Region onlineRegion = rs.getRegion(encodedName); 095 if (onlineRegion != null) { 096 LOG.warn("Received OPEN for the region:{}, which is already online", regionName); 097 // Just follow the old behavior, do we need to call reportRegionStateTransition? Maybe not? 098 // For normal case, it could happen that the rpc call to schedule this handler is succeeded, 099 // but before returning to master the connection is broken. And when master tries again, we 100 // have already finished the opening. For this case we do not need to call 101 // reportRegionStateTransition any more. 102 return; 103 } 104 Boolean previous = rs.getRegionsInTransitionInRS().putIfAbsent(encodedNameBytes, Boolean.TRUE); 105 if (previous != null) { 106 if (previous) { 107 // The region is opening and this maybe a retry on the rpc call, it is safe to ignore it. 108 LOG.info("Receiving OPEN for the region:{}, which we are already trying to OPEN" + 109 " - ignoring this new request for this region.", regionName); 110 } else { 111 // The region is closing. This is possible as we will update the region state to CLOSED when 112 // calling reportRegionStateTransition, so the HMaster will think the region is offline, 113 // before we actually close the region, as reportRegionStateTransition is part of the 114 // closing process. 115 long backoff = retryCounter.getBackoffTimeAndIncrementAttempts(); 116 LOG.info( 117 "Receiving OPEN for the region:{}, which we are trying to close, try again after {}ms", 118 regionName, backoff); 119 rs.getExecutorService().delayedSubmit(this, backoff, TimeUnit.MILLISECONDS); 120 } 121 return; 122 } 123 LOG.info("Open {}", regionName); 124 HRegion region; 125 try { 126 TableDescriptor htd = 127 tableDesc != null ? tableDesc : rs.getTableDescriptors().get(regionInfo.getTable()); 128 if (htd == null) { 129 throw new IOException("Missing table descriptor for " + regionName); 130 } 131 // pass null for the last parameter, which used to be a CancelableProgressable, as now the 132 // opening can not be interrupted by a close request any more. 133 region = HRegion.openHRegion(regionInfo, htd, rs.getWAL(regionInfo), rs.getConfiguration(), 134 rs, null); 135 } catch (IOException e) { 136 cleanUpAndReportFailure(e); 137 return; 138 } 139 rs.postOpenDeployTasks(new PostOpenDeployContext(region, openProcId, masterSystemTime)); 140 rs.addRegion(region); 141 LOG.info("Opened {}", regionName); 142 // Cache the open region procedure id after report region transition succeed. 143 rs.finishRegionProcedure(openProcId); 144 Boolean current = rs.getRegionsInTransitionInRS().remove(regionInfo.getEncodedNameAsBytes()); 145 if (current == null) { 146 // Should NEVER happen, but let's be paranoid. 147 LOG.error("Bad state: we've just opened a region that was NOT in transition. Region={}", 148 regionName); 149 } else if (!current) { 150 // Should NEVER happen, but let's be paranoid. 151 LOG.error("Bad state: we've just opened a region that was closing. Region={}", regionName); 152 } 153 } 154 155 @Override 156 protected void handleException(Throwable t) { 157 LOG.warn("Fatal error occurred while opening region {}, aborting...", 158 regionInfo.getRegionNameAsString(), t); 159 getServer().abort( 160 "Failed to open region " + regionInfo.getRegionNameAsString() + " and can not recover", t); 161 } 162 163 public static AssignRegionHandler create(HRegionServer server, RegionInfo regionInfo, 164 long openProcId, TableDescriptor tableDesc, long masterSystemTime) { 165 EventType eventType; 166 if (regionInfo.isMetaRegion()) { 167 eventType = EventType.M_RS_CLOSE_META; 168 } else if (regionInfo.getTable().isSystemTable() || 169 (tableDesc != null && tableDesc.getPriority() >= HConstants.ADMIN_QOS)) { 170 eventType = EventType.M_RS_OPEN_PRIORITY_REGION; 171 } else { 172 eventType = EventType.M_RS_OPEN_REGION; 173 } 174 return new AssignRegionHandler(server, regionInfo, openProcId, tableDesc, masterSystemTime, 175 eventType); 176 } 177}