/**
*
* 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.hadoop.hbase.master.normalizer;
import java.util.ArrayList;
import java.util.Collections;
import java.util.Comparator;
import java.util.List;
import org.apache.commons.logging.Log;
import org.apache.commons.logging.LogFactory;
import org.apache.hadoop.hbase.HBaseIOException;
import org.apache.hadoop.hbase.HRegionInfo;
import org.apache.hadoop.hbase.RegionLoad;
import org.apache.hadoop.hbase.ServerName;
import org.apache.hadoop.hbase.TableName;
import org.apache.hadoop.hbase.classification.InterfaceAudience;
import org.apache.hadoop.hbase.client.MasterSwitchType;
import org.apache.hadoop.hbase.master.MasterRpcServices;
import org.apache.hadoop.hbase.master.MasterServices;
import org.apache.hadoop.hbase.master.normalizer.NormalizationPlan.PlanType;
import org.apache.hadoop.hbase.shaded.protobuf.RequestConverter;
/**
* Simple implementation of region normalizer.
*
* Logic in use:
*
* <ol>
* <li> get all regions of a given table
* <li> get avg size S of each region (by total size of store files reported in RegionLoad)
* <li> If biggest region is bigger than S * 2, it is kindly requested to split,
* and normalization stops
* <li> Otherwise, two smallest region R1 and its smallest neighbor R2 are kindly requested
* to merge, if R1 + R1 < S, and normalization stops
* <li> Otherwise, no action is performed
* </ol>
* <p>
* Region sizes are coarse and approximate on the order of megabytes. Additionally,
* "empty" regions (less than 1MB, with the previous note) are not merged away. This
* is by design to prevent normalization from undoing the pre-splitting of a table.
*/
@InterfaceAudience.Private
public class SimpleRegionNormalizer implements RegionNormalizer {
private static final Log LOG = LogFactory.getLog(SimpleRegionNormalizer.class);
private static final int MIN_REGION_COUNT = 3;
private MasterServices masterServices;
private MasterRpcServices masterRpcServices;
private static long[] skippedCount = new long[NormalizationPlan.PlanType.values().length];
/**
* Set the master service.
* @param masterServices inject instance of MasterServices
*/
@Override
public void setMasterServices(MasterServices masterServices) {
this.masterServices = masterServices;
}
@Override
public void setMasterRpcServices(MasterRpcServices masterRpcServices) {
this.masterRpcServices = masterRpcServices;
}
@Override
public void planSkipped(HRegionInfo hri, PlanType type) {
skippedCount[type.ordinal()]++;
}
@Override
public long getSkippedCount(NormalizationPlan.PlanType type) {
return skippedCount[type.ordinal()];
}
// Comparator that gives higher priority to region Split plan
private Comparator<NormalizationPlan> planComparator =
new Comparator<NormalizationPlan>() {
@Override
public int compare(NormalizationPlan plan, NormalizationPlan plan2) {
if (plan instanceof SplitNormalizationPlan) {
return -1;
}
if (plan2 instanceof SplitNormalizationPlan) {
return 1;
}
return 0;
}
};
/**
* Computes next most "urgent" normalization action on the table.
* Action may be either a split, or a merge, or no action.
*
* @param table table to normalize
* @return normalization plan to execute
*/
@Override
public List<NormalizationPlan> computePlanForTable(TableName table) throws HBaseIOException {
if (table == null || table.isSystemTable()) {
LOG.debug("Normalization of system table " + table + " isn't allowed");
return null;
}
List<NormalizationPlan> plans = new ArrayList<>();
List<HRegionInfo> tableRegions = masterServices.getAssignmentManager().getRegionStates().
getRegionsOfTable(table);
//TODO: should we make min number of regions a config param?
if (tableRegions == null || tableRegions.size() < MIN_REGION_COUNT) {
int nrRegions = tableRegions == null ? 0 : tableRegions.size();
LOG.debug("Table " + table + " has " + nrRegions + " regions, required min number"
+ " of regions for normalizer to run is " + MIN_REGION_COUNT + ", not running normalizer");
return null;
}
LOG.debug("Computing normalization plan for table: " + table +
", number of regions: " + tableRegions.size());
long totalSizeMb = 0;
int acutalRegionCnt = 0;
for (int i = 0; i < tableRegions.size(); i++) {
HRegionInfo hri = tableRegions.get(i);
long regionSize = getRegionSize(hri);
if (regionSize > 0) {
acutalRegionCnt++;
totalSizeMb += regionSize;
}
}
double avgRegionSize = acutalRegionCnt == 0 ? 0 : totalSizeMb / (double) acutalRegionCnt;
LOG.debug("Table " + table + ", total aggregated regions size: " + totalSizeMb);
LOG.debug("Table " + table + ", average region size: " + avgRegionSize);
int candidateIdx = 0;
boolean splitEnabled = true, mergeEnabled = true;
try {
splitEnabled = masterRpcServices.isSplitOrMergeEnabled(null,
RequestConverter.buildIsSplitOrMergeEnabledRequest(MasterSwitchType.SPLIT)).getEnabled();
} catch (org.apache.hadoop.hbase.shaded.com.google.protobuf.ServiceException e) {
LOG.debug("Unable to determine whether split is enabled", e);
}
try {
mergeEnabled = masterRpcServices.isSplitOrMergeEnabled(null,
RequestConverter.buildIsSplitOrMergeEnabledRequest(MasterSwitchType.MERGE)).getEnabled();
} catch (org.apache.hadoop.hbase.shaded.com.google.protobuf.ServiceException e) {
LOG.debug("Unable to determine whether split is enabled", e);
}
while (candidateIdx < tableRegions.size()) {
HRegionInfo hri = tableRegions.get(candidateIdx);
long regionSize = getRegionSize(hri);
// if the region is > 2 times larger than average, we split it, split
// is more high priority normalization action than merge.
if (regionSize > 2 * avgRegionSize) {
if (splitEnabled) {
LOG.info("Table " + table + ", large region " + hri.getRegionNameAsString() + " has size "
+ regionSize + ", more than twice avg size, splitting");
plans.add(new SplitNormalizationPlan(hri, null));
}
} else {
if (candidateIdx == tableRegions.size()-1) {
break;
}
if (mergeEnabled) {
HRegionInfo hri2 = tableRegions.get(candidateIdx+1);
long regionSize2 = getRegionSize(hri2);
if (regionSize >= 0 && regionSize2 >= 0 && regionSize + regionSize2 < avgRegionSize) {
LOG.info("Table " + table + ", small region size: " + regionSize
+ " plus its neighbor size: " + regionSize2
+ ", less than the avg size " + avgRegionSize + ", merging them");
plans.add(new MergeNormalizationPlan(hri, hri2));
candidateIdx++;
}
}
}
candidateIdx++;
}
if (plans.isEmpty()) {
LOG.debug("No normalization needed, regions look good for table: " + table);
return null;
}
Collections.sort(plans, planComparator);
return plans;
}
private long getRegionSize(HRegionInfo hri) {
ServerName sn = masterServices.getAssignmentManager().getRegionStates().
getRegionServerOfRegion(hri);
RegionLoad regionLoad = masterServices.getServerManager().getLoad(sn).
getRegionsLoad().get(hri.getRegionName());
if (regionLoad == null) {
LOG.debug(hri.getRegionNameAsString() + " was not found in RegionsLoad");
return -1;
}
return regionLoad.getStorefileSizeMB();
}
}