[
https://issues.apache.org/jira/browse/HIVE-22977?focusedWorklogId=841206&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-841206
]
ASF GitHub Bot logged work on HIVE-22977:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 23/Jan/23 18:31
Start Date: 23/Jan/23 18:31
Worklog Time Spent: 10m
Work Description: SourabhBadhya commented on code in PR #3801:
URL: https://github.com/apache/hive/pull/3801#discussion_r1084406497
##########
ql/src/java/org/apache/hadoop/hive/ql/txn/compactor/MergeCompactor.java:
##########
@@ -0,0 +1,210 @@
+/*
+ * 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.hive.ql.txn.compactor;
+
+import org.apache.hadoop.fs.FileStatus;
+import org.apache.hadoop.fs.FileSystem;
+import org.apache.hadoop.fs.Path;
+import org.apache.hadoop.hive.common.ValidWriteIdList;
+import org.apache.hadoop.hive.conf.HiveConf;
+import org.apache.hadoop.hive.metastore.api.Partition;
+import org.apache.hadoop.hive.metastore.api.StorageDescriptor;
+import org.apache.hadoop.hive.metastore.api.Table;
+import org.apache.hadoop.hive.metastore.txn.CompactionInfo;
+import org.apache.hadoop.hive.ql.io.AcidDirectory;
+import org.apache.hadoop.hive.ql.io.AcidOutputFormat;
+import org.apache.hadoop.hive.ql.io.AcidUtils;
+import org.apache.hadoop.hive.ql.io.orc.OrcFile;
+import org.apache.hadoop.hive.ql.io.orc.Reader;
+import org.apache.hadoop.hive.ql.metadata.HiveException;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+import java.util.regex.Matcher;
+
+final class MergeCompactor extends QueryCompactor {
+
+ private static final Logger LOG =
LoggerFactory.getLogger(MergeCompactor.class.getName());
+
+ @Override
+ public boolean run(HiveConf hiveConf, Table table, Partition partition,
StorageDescriptor storageDescriptor,
+ ValidWriteIdList writeIds, CompactionInfo compactionInfo,
AcidDirectory dir) throws IOException, HiveException, InterruptedException {
+ if (isMergeCompaction(hiveConf, dir, writeIds, storageDescriptor)) {
+ // Only inserts happened, it is much more performant to merge the files
than running a query
+ Path outputDirPath = getCompactionOutputDirPath(hiveConf, writeIds,
+ compactionInfo.isMajorCompaction(), storageDescriptor);
+ try {
+ return mergeOrcFiles(hiveConf, compactionInfo.isMajorCompaction(),
+ dir, outputDirPath,
AcidUtils.isInsertOnlyTable(table.getParameters()));
+ } catch (Throwable t) {
+ // Error handling, just delete the output directory,
+ // and fall back to query based compaction.
+ FileSystem fs = outputDirPath.getFileSystem(hiveConf);
+ if (fs.exists(outputDirPath)) {
+ fs.delete(outputDirPath, true);
+ }
+ return false;
+ }
+ } else {
+ return false;
+ }
+ }
+
+ /**
+ * Returns whether merge compaction must be enabled or not.
+ * @param conf Hive configuration
+ * @param directory the directory to be scanned
+ * @param validWriteIdList list of valid write IDs
+ * @param storageDescriptor storage descriptor of the underlying table
+ * @return true, if merge compaction must be enabled
+ */
+ private boolean isMergeCompaction(HiveConf conf, AcidDirectory directory,
+ ValidWriteIdList validWriteIdList,
+ StorageDescriptor storageDescriptor) {
+ return conf.getBoolVar(HiveConf.ConfVars.HIVE_MERGE_COMPACTION_ENABLED)
+ &&
storageDescriptor.getOutputFormat().equalsIgnoreCase("org.apache.hadoop.hive.ql.io.orc.OrcOutputFormat")
+ && !hasDeleteOrAbortedDirectories(directory, validWriteIdList);
+ }
+
+ /**
+ * Scan a directory for delete deltas or aborted directories.
+ * @param directory the directory to be scanned
+ * @param validWriteIdList list of valid write IDs
+ * @return true, if delete or aborted directory found
+ */
+ private boolean hasDeleteOrAbortedDirectories(AcidDirectory directory,
ValidWriteIdList validWriteIdList) {
+ if (!directory.getCurrentDirectories().isEmpty()) {
+ final long minWriteId = validWriteIdList.getMinOpenWriteId() == null ? 1
: validWriteIdList.getMinOpenWriteId();
+ final long maxWriteId = validWriteIdList.getHighWatermark();
+ return directory.getCurrentDirectories().stream()
+ .filter(AcidUtils.ParsedDeltaLight::isDeleteDelta)
+ .filter(delta -> delta.getMinWriteId() >= minWriteId)
+ .anyMatch(delta -> delta.getMaxWriteId() <= maxWriteId) ||
!directory.getAbortedDirectories().isEmpty();
+ }
+ return true;
+ }
+
+ /**
+ * Collect the list of all bucket file paths, which belong to the same
bucket Id. This method scans all the base
+ * and delta dirs.
+ * @param conf hive configuration, must be not null
+ * @param dir the root directory of delta dirs
+ * @param includeBaseDir true, if the base directory should be scanned
+ * @param isMm
+ * @return map of bucket ID -> bucket files
+ * @throws IOException an error happened during the reading of the
directory/bucket file
+ */
+ private Map<Integer, List<Reader>> matchBucketIdToBucketFiles(HiveConf conf,
AcidDirectory dir,
+ boolean
includeBaseDir, boolean isMm) throws IOException {
+ Map<Integer, List<Reader>> result = new HashMap<>();
+ if (includeBaseDir && dir.getBaseDirectory() != null) {
+ getBucketFiles(conf, dir.getBaseDirectory(), isMm, result);
+ }
+ for (AcidUtils.ParsedDelta deltaDir : dir.getCurrentDirectories()) {
+ Path deltaDirPath = deltaDir.getPath();
+ getBucketFiles(conf, deltaDirPath, isMm, result);
Review Comment:
Refactored it. However, it is not as simple as - `result.putAll`.
I think the earlier way was more efficient. We are making extra iterations
to include the bucket files from the delta dir to the overall result.
##########
ql/src/java/org/apache/hadoop/hive/ql/txn/compactor/MergeCompactor.java:
##########
@@ -0,0 +1,210 @@
+/*
+ * 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.hive.ql.txn.compactor;
+
+import org.apache.hadoop.fs.FileStatus;
+import org.apache.hadoop.fs.FileSystem;
+import org.apache.hadoop.fs.Path;
+import org.apache.hadoop.hive.common.ValidWriteIdList;
+import org.apache.hadoop.hive.conf.HiveConf;
+import org.apache.hadoop.hive.metastore.api.Partition;
+import org.apache.hadoop.hive.metastore.api.StorageDescriptor;
+import org.apache.hadoop.hive.metastore.api.Table;
+import org.apache.hadoop.hive.metastore.txn.CompactionInfo;
+import org.apache.hadoop.hive.ql.io.AcidDirectory;
+import org.apache.hadoop.hive.ql.io.AcidOutputFormat;
+import org.apache.hadoop.hive.ql.io.AcidUtils;
+import org.apache.hadoop.hive.ql.io.orc.OrcFile;
+import org.apache.hadoop.hive.ql.io.orc.Reader;
+import org.apache.hadoop.hive.ql.metadata.HiveException;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+import java.util.regex.Matcher;
+
+final class MergeCompactor extends QueryCompactor {
+
+ private static final Logger LOG =
LoggerFactory.getLogger(MergeCompactor.class.getName());
+
+ @Override
+ public boolean run(HiveConf hiveConf, Table table, Partition partition,
StorageDescriptor storageDescriptor,
+ ValidWriteIdList writeIds, CompactionInfo compactionInfo,
AcidDirectory dir) throws IOException, HiveException, InterruptedException {
+ if (isMergeCompaction(hiveConf, dir, writeIds, storageDescriptor)) {
+ // Only inserts happened, it is much more performant to merge the files
than running a query
+ Path outputDirPath = getCompactionOutputDirPath(hiveConf, writeIds,
+ compactionInfo.isMajorCompaction(), storageDescriptor);
+ try {
+ return mergeOrcFiles(hiveConf, compactionInfo.isMajorCompaction(),
+ dir, outputDirPath,
AcidUtils.isInsertOnlyTable(table.getParameters()));
+ } catch (Throwable t) {
+ // Error handling, just delete the output directory,
+ // and fall back to query based compaction.
+ FileSystem fs = outputDirPath.getFileSystem(hiveConf);
+ if (fs.exists(outputDirPath)) {
+ fs.delete(outputDirPath, true);
+ }
+ return false;
+ }
+ } else {
+ return false;
+ }
+ }
+
+ /**
+ * Returns whether merge compaction must be enabled or not.
+ * @param conf Hive configuration
+ * @param directory the directory to be scanned
+ * @param validWriteIdList list of valid write IDs
+ * @param storageDescriptor storage descriptor of the underlying table
+ * @return true, if merge compaction must be enabled
+ */
+ private boolean isMergeCompaction(HiveConf conf, AcidDirectory directory,
+ ValidWriteIdList validWriteIdList,
+ StorageDescriptor storageDescriptor) {
+ return conf.getBoolVar(HiveConf.ConfVars.HIVE_MERGE_COMPACTION_ENABLED)
+ &&
storageDescriptor.getOutputFormat().equalsIgnoreCase("org.apache.hadoop.hive.ql.io.orc.OrcOutputFormat")
+ && !hasDeleteOrAbortedDirectories(directory, validWriteIdList);
+ }
+
+ /**
+ * Scan a directory for delete deltas or aborted directories.
+ * @param directory the directory to be scanned
+ * @param validWriteIdList list of valid write IDs
+ * @return true, if delete or aborted directory found
+ */
+ private boolean hasDeleteOrAbortedDirectories(AcidDirectory directory,
ValidWriteIdList validWriteIdList) {
+ if (!directory.getCurrentDirectories().isEmpty()) {
+ final long minWriteId = validWriteIdList.getMinOpenWriteId() == null ? 1
: validWriteIdList.getMinOpenWriteId();
+ final long maxWriteId = validWriteIdList.getHighWatermark();
+ return directory.getCurrentDirectories().stream()
+ .filter(AcidUtils.ParsedDeltaLight::isDeleteDelta)
+ .filter(delta -> delta.getMinWriteId() >= minWriteId)
+ .anyMatch(delta -> delta.getMaxWriteId() <= maxWriteId) ||
!directory.getAbortedDirectories().isEmpty();
+ }
+ return true;
+ }
+
+ /**
+ * Collect the list of all bucket file paths, which belong to the same
bucket Id. This method scans all the base
+ * and delta dirs.
+ * @param conf hive configuration, must be not null
+ * @param dir the root directory of delta dirs
+ * @param includeBaseDir true, if the base directory should be scanned
+ * @param isMm
+ * @return map of bucket ID -> bucket files
+ * @throws IOException an error happened during the reading of the
directory/bucket file
+ */
+ private Map<Integer, List<Reader>> matchBucketIdToBucketFiles(HiveConf conf,
AcidDirectory dir,
+ boolean
includeBaseDir, boolean isMm) throws IOException {
+ Map<Integer, List<Reader>> result = new HashMap<>();
+ if (includeBaseDir && dir.getBaseDirectory() != null) {
+ getBucketFiles(conf, dir.getBaseDirectory(), isMm, result);
+ }
+ for (AcidUtils.ParsedDelta deltaDir : dir.getCurrentDirectories()) {
+ Path deltaDirPath = deltaDir.getPath();
+ getBucketFiles(conf, deltaDirPath, isMm, result);
+ }
+ return result;
+ }
+
+ /**
+ * Collect the list of all bucket file paths, which belong to the same
bucket Id. This method checks only one
+ * directory.
+ * @param conf hive configuration, must be not null
+ * @param dirPath the directory to be scanned.
+ * @param isMm collect bucket files fron insert only directories
+ * @param bucketIdToBucketFilePath the result of the scan
+ * @throws IOException an error happened during the reading of the
directory/bucket file
+ */
+ private void getBucketFiles(HiveConf conf, Path dirPath, boolean isMm,
Map<Integer, List<Reader>> bucketIdToBucketFilePath) throws IOException {
+ FileSystem fs = dirPath.getFileSystem(conf);
+ FileStatus[] fileStatuses =
+ fs.listStatus(dirPath, isMm ? AcidUtils.originalBucketFilter :
AcidUtils.bucketFileFilter);
+ for (FileStatus f : fileStatuses) {
+ final Path fPath = f.getPath();
+ Matcher matcher = isMm ? AcidUtils.LEGACY_BUCKET_DIGIT_PATTERN
+ .matcher(fPath.getName()) :
AcidUtils.BUCKET_PATTERN.matcher(fPath.getName());
+ if (!matcher.find()) {
+ String errorMessage = String
+ .format("Found a bucket file which did not match the bucket
pattern! %s Matcher=%s", fPath,
+ matcher);
+ LOG.error(errorMessage);
+ throw new IllegalArgumentException(errorMessage);
+ }
+ int bucketNum = matcher.groupCount() > 0 ?
Integer.parseInt(matcher.group(1)) : Integer.parseInt(matcher.group());
+ bucketIdToBucketFilePath.computeIfAbsent(bucketNum, ArrayList::new);
+ Reader reader = OrcFile.createReader(fs, fPath);
+ bucketIdToBucketFilePath.computeIfPresent(bucketNum, (k, v) ->
v).add(reader);
+ }
+ }
+
+ /**
+ * Generate output path for compaction. This can be used to generate delta
or base directories.
+ * @param conf hive configuration, must be non-null
+ * @param writeIds list of valid write IDs
+ * @param isBaseDir if base directory path should be generated
+ * @param sd the resolved storadge descriptor
+ * @return output path, always non-null
+ */
+ private Path getCompactionOutputDirPath(HiveConf conf, ValidWriteIdList
writeIds, boolean isBaseDir,
+ StorageDescriptor sd) {
+ long minOpenWriteId = writeIds.getMinOpenWriteId() == null ? 1 :
writeIds.getMinOpenWriteId();
+ long highWatermark = writeIds.getHighWatermark();
+ long compactorTxnId = Compactor.getCompactorTxnId(conf);
+ AcidOutputFormat.Options options = new
AcidOutputFormat.Options(conf).writingBase(isBaseDir)
+
.writingDeleteDelta(false).isCompressed(false).minimumWriteId(minOpenWriteId)
+
.maximumWriteId(highWatermark).statementId(-1).visibilityTxnId(compactorTxnId);
+ return AcidUtils.baseOrDeltaSubdirPath(new Path(sd.getLocation()),
options);
+ }
+
+ /**
+ * Merge ORC files from base/delta directories. If the directories contains
multiple buckets, the result will also
+ * contain the same amount.
+ * @param conf hive configuration
+ * @param includeBaseDir if base directory should be scanned for orc files
+ * @param dir the root directory of the table/partition
+ * @param outputDirPath the result directory path
+ * @param isMm merge orc files from insert only tables
+ * @throws IOException error occurred during file operation
+ */
+ private boolean mergeOrcFiles(HiveConf conf, boolean includeBaseDir,
AcidDirectory dir,
Review Comment:
I have renamed it to mergeFiles. Currently we support only to ORC full acid
and ORC insert-only tables.
We could think of other file formats as well. File formats which supports
reading in batches for all datatypes is probably optimal for merge compaction
(in terms of performance). However, I will keep it as a scope for another Jira.
Issue Time Tracking
-------------------
Worklog Id: (was: 841206)
Time Spent: 6h 50m (was: 6h 40m)
> Merge delta files instead of running a query in major/minor compaction
> ----------------------------------------------------------------------
>
> Key: HIVE-22977
> URL: https://issues.apache.org/jira/browse/HIVE-22977
> Project: Hive
> Issue Type: Improvement
> Reporter: László Pintér
> Assignee: Sourabh Badhya
> Priority: Major
> Labels: pull-request-available
> Attachments: HIVE-22977.01.patch, HIVE-22977.02.patch
>
> Time Spent: 6h 50m
> Remaining Estimate: 0h
>
> [Compaction Optimiziation]
> We should analyse the possibility to move a delta file instead of running a
> major/minor compaction query.
> Please consider the following use cases:
> - full acid table but only insert queries were run. This means that no
> delete delta directories were created. Is it possible to merge the delta
> directory contents without running a compaction query?
> - full acid table, initiating queries through the streaming API. If there
> are no abort transactions during the streaming, is it possible to merge the
> delta directory contents without running a compaction query?
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
