You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by GitBox <gi...@apache.org> on 2021/10/24 13:55:59 UTC
[GitHub] [hbase] Apache9 commented on a change in pull request #3786: HBASE-26271: Cleanup the broken store files under data directory
Apache9 commented on a change in pull request #3786:
URL: https://github.com/apache/hbase/pull/3786#discussion_r735122599
##########
File path: hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/storefiletracker/StoreFileTrackerBase.java
##########
@@ -176,7 +176,7 @@ public final StoreFileWriter createWriter(CreateStoreFileWriterParams params) th
* Whether the implementation of this tracker requires you to write to temp directory first, i.e,
* does not allow broken store files under the actual data directory.
*/
- protected abstract boolean requireWritingToTmpDirFirst();
+ public abstract boolean requireWritingToTmpDirFirst();
Review comment:
Just remove this line if we move this method up to StoreFileTracker.
##########
File path: hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/storefiletracker/StoreFileTracker.java
##########
@@ -88,4 +88,10 @@ void replace(Collection<StoreFileInfo> compactedFiles, Collection<StoreFileInfo>
* @param builder The table descriptor builder for the given table.
*/
TableDescriptorBuilder updateWithTrackerConfigs(TableDescriptorBuilder builder);
+
+ /**
+ * Whether the implementation of this tracker requires you to write to temp directory first, i.e,
+ * does not allow broken store files under the actual data directory.
+ */
+ public boolean requireWritingToTmpDirFirst();
Review comment:
nits: StoreFileTracker is an interface so we do not need the public modifier here.
##########
File path: hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/compactions/Compactor.java
##########
@@ -537,4 +547,17 @@ protected InternalScanner createScanner(HStore store, ScanInfo scanInfo,
return new StoreScanner(store, scanInfo, scanners, smallestReadPoint, earliestPutTs,
dropDeletesFromRow, dropDeletesToRow);
}
+
+ public List<Path> getCompactionTargets(){
+ if (writer == null){
Review comment:
OK, so this is the trick here, in general I think the approach is fine, after committing we will set writer to null. But I think we need to make sure that we only set writer to null after we finish committing these files to store file tracker. In the current implementation, I do not think it is the case, at least in the below DateTieredCompactor, we set writer to null directly after we commit(it is just a close actually) the writers, we will then use the returned files to update the store file tracker.
##########
File path: hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/FileBasedStoreFileCleaner.java
##########
@@ -0,0 +1,191 @@
+/**
+ * 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.hbase.regionserver;
+
+import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.fs.FileStatus;
+import org.apache.hadoop.fs.Path;
+import org.apache.hadoop.hbase.ScheduledChore;
+import org.apache.hadoop.hbase.Stoppable;
+import org.apache.hadoop.hbase.io.HFileLink;
+import org.apache.hadoop.hbase.util.EnvironmentEdgeManager;
+import org.apache.hadoop.ipc.RemoteException;
+import org.apache.yetus.audience.InterfaceAudience;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.Arrays;
+import java.util.List;
+import java.util.concurrent.atomic.AtomicBoolean;
+import java.util.concurrent.atomic.AtomicLong;
+
+/**
+ * This Chore, every time it runs, will clear the unsused HFiles in the data
+ * folder.
+ */
+@InterfaceAudience.Private public class FileBasedStoreFileCleaner extends ScheduledChore {
Review comment:
Let's not name it 'FileBased'. I think it should be more general, maybe 'BrokenStoreFileCleaner'.
For now, the only condition is whether we will write to the data directory directly. No matter what is the actual store file tracker implementation, if it writes to the data directory directly, then we need this cleaner, otherwise we do not.
##########
File path: hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/FileBasedStoreFileCleaner.java
##########
@@ -0,0 +1,191 @@
+/**
+ * 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.hbase.regionserver;
+
+import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.fs.FileStatus;
+import org.apache.hadoop.fs.Path;
+import org.apache.hadoop.hbase.ScheduledChore;
+import org.apache.hadoop.hbase.Stoppable;
+import org.apache.hadoop.hbase.io.HFileLink;
+import org.apache.hadoop.hbase.util.EnvironmentEdgeManager;
+import org.apache.hadoop.ipc.RemoteException;
+import org.apache.yetus.audience.InterfaceAudience;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.Arrays;
+import java.util.List;
+import java.util.concurrent.atomic.AtomicBoolean;
+import java.util.concurrent.atomic.AtomicLong;
+
+/**
+ * This Chore, every time it runs, will clear the unsused HFiles in the data
+ * folder.
+ */
+@InterfaceAudience.Private public class FileBasedStoreFileCleaner extends ScheduledChore {
+ private static final Logger LOG = LoggerFactory.getLogger(FileBasedStoreFileCleaner.class);
+ public static final String FILEBASED_STOREFILE_CLEANER_ENABLED =
+ "hbase.region.filebased.storefilecleaner.enabled";
+ public static final boolean DEFAULT_FILEBASED_STOREFILE_CLEANER_ENABLED = false;
+ public static final String FILEBASED_STOREFILE_CLEANER_TTL =
+ "hbase.region.filebased.storefilecleaner.ttl";
+ public static final long DEFAULT_FILEBASED_STOREFILE_CLEANER_TTL = 1000 * 60 * 60 * 12; //12h
+ public static final String FILEBASED_STOREFILE_CLEANER_DELAY =
+ "hbase.region.filebased.storefilecleaner.delay";
+ public static final int DEFAULT_FILEBASED_STOREFILE_CLEANER_DELAY = 1000 * 60 * 60 * 2; //2h
+ public static final String FILEBASED_STOREFILE_CLEANER_DELAY_JITTER =
+ "hbase.region.filebased.storefilecleaner.delay.jitter";
+ public static final double DEFAULT_FILEBASED_STOREFILE_CLEANER_DELAY_JITTER = 0.25D;
+ public static final String FILEBASED_STOREFILE_CLEANER_PERIOD =
+ "hbase.region.filebased.storefilecleaner.period";
+ public static final int DEFAULT_FILEBASED_STOREFILE_CLEANER_PERIOD = 1000 * 60 * 60 * 6; //6h
+
+ private HRegionServer regionServer;
+ private final AtomicBoolean enabled = new AtomicBoolean(true);
+ private long ttl;
+
+ public FileBasedStoreFileCleaner(final int delay, final int period, final Stoppable stopper, Configuration conf,
+ HRegionServer regionServer) {
+ super("FileBasedStoreFileCleaner", stopper, period, delay);
+ this.regionServer = regionServer;
+ setEnabled(conf.getBoolean(FILEBASED_STOREFILE_CLEANER_ENABLED, DEFAULT_FILEBASED_STOREFILE_CLEANER_ENABLED));
+ ttl = conf.getLong(FILEBASED_STOREFILE_CLEANER_TTL, DEFAULT_FILEBASED_STOREFILE_CLEANER_TTL);
+ }
+
+ public boolean setEnabled(final boolean enabled) {
+ return this.enabled.getAndSet(enabled);
+ }
+
+ public boolean getEnabled() {
+ return this.enabled.get();
+ }
+
+ @InterfaceAudience.Private
+ @Override public void chore() {
+ if (getEnabled()) {
+ long start = EnvironmentEdgeManager.currentTime();
+ AtomicLong deletedFiles = new AtomicLong(0);
+ AtomicLong failedDeletes = new AtomicLong(0);
+ for (HRegion region : regionServer.getRegions()) {
+ for (HStore store : region.getStores()) {
+ //only clean do cleanup in store using file based storefile tracking
+ if (store.getStoreEngine().requireWritingToTmpDirFirst()) {
+ continue;
+ }
+ Path storePath =
+ new Path(region.getRegionFileSystem().getRegionDir(), store.getColumnFamilyName());
+
+ try {
+ List<FileStatus> fsStoreFiles = Arrays.asList(region.getRegionFileSystem().fs.listStatus(storePath));
+ fsStoreFiles.forEach(file -> cleanFileIfNeeded(file, store, deletedFiles, failedDeletes));
+ } catch (IOException e) {
+ LOG.warn("Failed to list files in {}, cleanup is skipped there",storePath);
+ continue;
+ }
+ }
+ }
+ LOG.debug(
+ "FileBasedStoreFileCleaner on {} run for: {}ms. It deleted {} files and tried but failed to delete {}",
+ regionServer.getServerName().getServerName(), EnvironmentEdgeManager.currentTime() - start,
+ deletedFiles.get(), failedDeletes.get());
+ } else {
+ LOG.trace("File based storefile Cleaner chore disabled! Not cleaning.");
+ }
+ }
+
+ private void cleanFileIfNeeded(FileStatus file, HStore store,
+ AtomicLong deletedFiles, AtomicLong failedDeletes) {
+ if(file.isDirectory()){
+ LOG.trace("This is a Directory {}, skip cleanup", file.getPath());
+ return;
+ }
+
+ if(!validate(file.getPath())){
+ LOG.trace("Invalid file {}, skip cleanup", file.getPath());
+ return;
+ }
+
+ if(!isOldEnough(file)){
+ LOG.trace("Fresh file {}, skip cleanup", file.getPath());
+ return;
+ }
+
+ if(isActiveStorefile(file, store)){
+ LOG.trace("Actively used storefile file {}, skip cleanup", file.getPath());
+ return;
+ }
+
+ if(isCompactedFile(file, store)){
+ LOG.trace("Cleanup is done by a different chore for file {}, skip cleanup", file.getPath());
+ return;
+ }
+
+ if(isCompactingFile(file, store)){
+ LOG.trace("The file is the result of an ongoing compaction {}, skip cleanup", file.getPath());
+ return;
+ }
+
+ deleteFile(file, store, deletedFiles, failedDeletes);
+ }
+
+ private boolean isCompactingFile(FileStatus file, HStore store) {
Review comment:
I suppose we should move the below isXXXFile to HRegion or HStore, with the protection of some locks. Otherwise we may have race, or at least, we depend on a very flaky order of the testing of each condition.
And do we really need to test isCompactingFile here? The compactedFile will not be tracked but compacting files should always be tracked?
##########
File path: hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/FileBasedStoreFileCleaner.java
##########
@@ -0,0 +1,191 @@
+/**
+ * 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.hbase.regionserver;
+
+import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.fs.FileStatus;
+import org.apache.hadoop.fs.Path;
+import org.apache.hadoop.hbase.ScheduledChore;
+import org.apache.hadoop.hbase.Stoppable;
+import org.apache.hadoop.hbase.io.HFileLink;
+import org.apache.hadoop.hbase.util.EnvironmentEdgeManager;
+import org.apache.hadoop.ipc.RemoteException;
+import org.apache.yetus.audience.InterfaceAudience;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.Arrays;
+import java.util.List;
+import java.util.concurrent.atomic.AtomicBoolean;
+import java.util.concurrent.atomic.AtomicLong;
+
+/**
+ * This Chore, every time it runs, will clear the unsused HFiles in the data
+ * folder.
+ */
+@InterfaceAudience.Private public class FileBasedStoreFileCleaner extends ScheduledChore {
+ private static final Logger LOG = LoggerFactory.getLogger(FileBasedStoreFileCleaner.class);
+ public static final String FILEBASED_STOREFILE_CLEANER_ENABLED =
+ "hbase.region.filebased.storefilecleaner.enabled";
+ public static final boolean DEFAULT_FILEBASED_STOREFILE_CLEANER_ENABLED = false;
+ public static final String FILEBASED_STOREFILE_CLEANER_TTL =
+ "hbase.region.filebased.storefilecleaner.ttl";
+ public static final long DEFAULT_FILEBASED_STOREFILE_CLEANER_TTL = 1000 * 60 * 60 * 12; //12h
+ public static final String FILEBASED_STOREFILE_CLEANER_DELAY =
+ "hbase.region.filebased.storefilecleaner.delay";
+ public static final int DEFAULT_FILEBASED_STOREFILE_CLEANER_DELAY = 1000 * 60 * 60 * 2; //2h
+ public static final String FILEBASED_STOREFILE_CLEANER_DELAY_JITTER =
+ "hbase.region.filebased.storefilecleaner.delay.jitter";
+ public static final double DEFAULT_FILEBASED_STOREFILE_CLEANER_DELAY_JITTER = 0.25D;
+ public static final String FILEBASED_STOREFILE_CLEANER_PERIOD =
+ "hbase.region.filebased.storefilecleaner.period";
+ public static final int DEFAULT_FILEBASED_STOREFILE_CLEANER_PERIOD = 1000 * 60 * 60 * 6; //6h
+
+ private HRegionServer regionServer;
+ private final AtomicBoolean enabled = new AtomicBoolean(true);
+ private long ttl;
+
+ public FileBasedStoreFileCleaner(final int delay, final int period, final Stoppable stopper, Configuration conf,
+ HRegionServer regionServer) {
+ super("FileBasedStoreFileCleaner", stopper, period, delay);
+ this.regionServer = regionServer;
+ setEnabled(conf.getBoolean(FILEBASED_STOREFILE_CLEANER_ENABLED, DEFAULT_FILEBASED_STOREFILE_CLEANER_ENABLED));
+ ttl = conf.getLong(FILEBASED_STOREFILE_CLEANER_TTL, DEFAULT_FILEBASED_STOREFILE_CLEANER_TTL);
+ }
+
+ public boolean setEnabled(final boolean enabled) {
+ return this.enabled.getAndSet(enabled);
+ }
+
+ public boolean getEnabled() {
+ return this.enabled.get();
+ }
+
+ @InterfaceAudience.Private
+ @Override public void chore() {
+ if (getEnabled()) {
+ long start = EnvironmentEdgeManager.currentTime();
+ AtomicLong deletedFiles = new AtomicLong(0);
+ AtomicLong failedDeletes = new AtomicLong(0);
+ for (HRegion region : regionServer.getRegions()) {
+ for (HStore store : region.getStores()) {
+ //only clean do cleanup in store using file based storefile tracking
+ if (store.getStoreEngine().requireWritingToTmpDirFirst()) {
Review comment:
OK, good, this is exactly what I expected. So the only problem is the naming. Let's find a better name.
--
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: issues-unsubscribe@hbase.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org