HIVE-28755: Statistics Management Task #6438
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,277 @@ | ||
| /* | ||
| * 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.metastore; | ||
|
|
||
| import java.util.ArrayList; | ||
| import java.util.Collections; | ||
| import java.util.LinkedHashMap; | ||
| import java.util.List; | ||
| import java.util.Map; | ||
| import java.util.concurrent.TimeUnit; | ||
|
DanielZhu58 marked this conversation as resolved.
|
||
| import java.util.concurrent.locks.Lock; | ||
| import java.util.concurrent.locks.ReentrantLock; | ||
|
|
||
| import org.apache.hadoop.conf.Configuration; | ||
| import org.apache.hadoop.hive.metastore.conf.MetastoreConf; | ||
| import org.apache.hadoop.hive.metastore.model.MPartitionColumnStatistics; | ||
| import org.apache.hadoop.hive.metastore.model.MTableColumnStatistics; | ||
| import org.slf4j.Logger; | ||
| import org.slf4j.LoggerFactory; | ||
|
|
||
| import javax.jdo.PersistenceManager; | ||
| import javax.jdo.Query; | ||
|
|
||
| /** | ||
| * Statistics management task responsible for periodic auto-deletion of table and partition column | ||
| * statistics based on a configured retention interval. | ||
| * | ||
| * <p>When {@code metastore.column.statistics.auto.deletion} is enabled, this task scans | ||
| * {@code TAB_COL_STATS} and {@code PART_COL_STATS} for rows whose {@code lastAnalyzed} timestamp | ||
| * is older than {@code metastore.column.statistics.retention.period}, and deletes them. | ||
| * Individual tables may opt out by setting the table property | ||
| * {@value #STATISTICS_AUTO_DELETION_EXCLUDE_TBLPROPERTY} to {@code "true"}. | ||
| */ | ||
| public class StatisticsManagementTask extends ObjectStore implements MetastoreTaskThread { | ||
|
|
||
| private static final Logger LOG = LoggerFactory.getLogger(StatisticsManagementTask.class); | ||
|
|
||
| /** | ||
| * Table property key that, when set to {@code "true"} on a table, excludes it from automatic | ||
| * statistics deletion regardless of the global retention setting. | ||
| */ | ||
| public static final String STATISTICS_AUTO_DELETION_EXCLUDE_TBLPROPERTY = | ||
| "statistics.auto.deletion.exclude"; | ||
|
Member
There was a problem hiding this comment. metastore.statistics.auto.deletion.exclude ?
Contributor
Author
There was a problem hiding this comment. That's a good point and I agree that database-level exclusion would be more convenient when users want to exclude all tables in a database. However, that's a very rare case. The Jira is created to delete the stale column stats, the exclusion is just for the users to prevent. Since this feature usually delete the column stats created one year ago, it is rare that the users want to exclude a certain table out of auto deletion. We can defer that to a follow-up Jira later.
Member
There was a problem hiding this comment. The "one year ago" is the default value, and it allows the user to set it based on their workload, and it wouldn't be good to exclude this feature per table/partition, in case having thousands of tables or partitions. Can you file the jira and link to this one please?
Contributor
Author
There was a problem hiding this comment. Acknowledged. Here is the new Jira link: https://issues.apache.org/jira/browse/HIVE-29608 |
||
|
|
||
| /** Separator used when building composite map keys; chosen to be safe in HMS identifiers. */ | ||
| private static final String KEY_SEP = "\0"; | ||
|
|
||
| private static final Lock LOCK = new ReentrantLock(); | ||
|
Member
There was a problem hiding this comment. nit: we don't need the
Contributor
Author
There was a problem hiding this comment. Acknowledged. I will keep this for now. |
||
|
|
||
| @Override | ||
| public long runFrequency(TimeUnit unit) { | ||
| // when frequency=0, the auto deletion task is not being run | ||
| if (MetastoreConf.getTimeVar(conf, MetastoreConf.ConfVars.COLUMN_STATISTICS_RETENTION_PERIOD, | ||
| TimeUnit.MILLISECONDS) <= 0) { | ||
| return 0; | ||
| } | ||
| return MetastoreConf.getTimeVar(conf, | ||
| MetastoreConf.ConfVars.COLUMN_STATISTICS_MANAGEMENT_TASK_FREQUENCY, unit); | ||
| } | ||
|
|
||
| @Override | ||
| public void setConf(Configuration configuration) { | ||
| this.conf = new Configuration(configuration); | ||
| super.setConf(this.conf); | ||
| } | ||
|
|
||
| @Override | ||
| public Configuration getConf() { | ||
| return conf; | ||
| } | ||
|
|
||
| @Override | ||
| public void run() { | ||
| LOG.debug("Auto statistics deletion started. Cleaning up table/partition column statistics" | ||
| + " over the retention period."); | ||
| long retentionMillis = MetastoreConf.getTimeVar( | ||
| conf, MetastoreConf.ConfVars.COLUMN_STATISTICS_RETENTION_PERIOD, TimeUnit.MILLISECONDS); | ||
| if (retentionMillis <= 0) { | ||
| LOG.info("Statistics auto deletion is set to off currently."); | ||
| return; | ||
| } | ||
| if (!LOCK.tryLock()) { | ||
| return; | ||
| } | ||
| try { | ||
| long now = System.currentTimeMillis(); | ||
| long lastAnalyzedThreshold = (now - retentionMillis) / 1000; | ||
| List<Object[]> expiredTblRows; | ||
| List<Object[]> expiredPartRows; | ||
| boolean committed = false; | ||
| openTransaction(); | ||
| try { | ||
| PersistenceManager pm = getPersistenceManager(); | ||
| expiredTblRows = collectExpiredTableColStats(pm, lastAnalyzedThreshold); | ||
| expiredPartRows = collectExpiredPartitionColStats(pm, lastAnalyzedThreshold); | ||
| committed = commitTransaction(); | ||
| } finally { | ||
| if (!committed) { | ||
| rollbackTransaction(); | ||
| } | ||
| } | ||
| deleteExpiredTableColStats(expiredTblRows); | ||
| deleteExpiredPartitionColStats(expiredPartRows); | ||
| } catch (Exception e) { | ||
| LOG.error("Error during statistics auto deletion", e); | ||
| } finally { | ||
| LOCK.unlock(); | ||
| } | ||
| } | ||
|
|
||
| /** | ||
| * Queries {@code TAB_COL_STATS} for rows whose {@code lastAnalyzed} is older than the given | ||
| * threshold. Results are copied into an {@link ArrayList} so they remain accessible after | ||
| * the enclosing JDO transaction is committed. | ||
| * | ||
| * @param pm the JDO persistence manager to use for the query | ||
| * @param lastAnalyzedThreshold epoch seconds; rows with lastAnalyzed below this value are expired | ||
| * @return list of projected rows: [catName, dbName, tblName, colName] | ||
| * @throws Exception if the JDO query fails | ||
| */ | ||
| private List<Object[]> collectExpiredTableColStats(PersistenceManager pm, | ||
| long lastAnalyzedThreshold) throws Exception { | ||
| try (Query tblQuery = pm.newQuery(MTableColumnStatistics.class)) { | ||
| tblQuery.setFilter("lastAnalyzed < threshold && engine == \"hive\""); | ||
| tblQuery.declareParameters("long threshold"); | ||
|
dengzhhu653 marked this conversation as resolved.
|
||
| tblQuery.setRange(0, 1000); | ||
| tblQuery.setResult( | ||
| "table.database.catalogName, " | ||
| + "table.database.name, " | ||
| + "table.tableName, " | ||
| + "colName, " | ||
| + "table.parameters.get(\"" + STATISTICS_AUTO_DELETION_EXCLUDE_TBLPROPERTY + "\")"); | ||
| @SuppressWarnings("unchecked") | ||
| List<Object[]> rows = (List<Object[]>) tblQuery.execute(lastAnalyzedThreshold); | ||
| return new ArrayList<>(rows); | ||
| } | ||
|
Comment on lines
+139
to
+152
Contributor
Author
There was a problem hiding this comment. Acknowledged. I added a filter which only collect the hive table and hive partitions. We do not want to touch any impala/spark tables and partitions. |
||
| } | ||
|
|
||
| /** | ||
| * Groups expired table-level column stats by {@code (catName, dbName, tblName)}, then deletes | ||
| * all expired columns for each table in a single {@link RawStore} call and a dedicated | ||
| * transaction. This avoids duplicate listener events for the same table. | ||
| * Tables whose exclude property is set to {@code "true"} are skipped entirely. | ||
| * | ||
| * @param expiredTblRows projected rows from {@link #collectExpiredTableColStats} | ||
| * @throws Exception if a delete operation fails | ||
| */ | ||
| private void deleteExpiredTableColStats(List<Object[]> expiredTblRows) throws Exception { | ||
| // key: catName + SEP + dbName + SEP + tblName, value: list of expired colNames | ||
| Map<String, List<String>> tblToColsMap = new LinkedHashMap<>(); | ||
| // keep a parallel map to reconstruct the key parts when issuing deletes | ||
| Map<String, String[]> keyToCoords = new LinkedHashMap<>(); | ||
|
|
||
| for (Object[] row : expiredTblRows) { | ||
| String catName = (String) row[0]; | ||
| String dbName = (String) row[1]; | ||
| String tblName = (String) row[2]; | ||
| String colName = (String) row[3]; | ||
| String excludeVal = (String) row[4]; | ||
| if (Boolean.parseBoolean(excludeVal)) { | ||
| LOG.info("Skipping auto deletion of table stats for {}.{} due to exclude property.", | ||
| dbName, tblName); | ||
| continue; | ||
| } | ||
| String key = catName + KEY_SEP + dbName + KEY_SEP + tblName; | ||
| tblToColsMap.computeIfAbsent(key, k -> new ArrayList<>()).add(colName); | ||
| keyToCoords.putIfAbsent(key, new String[]{catName, dbName, tblName}); | ||
| } | ||
|
|
||
| // one transaction per table, delete all expired columns in a single call | ||
| for (Map.Entry<String, List<String>> entry : tblToColsMap.entrySet()) { | ||
| String[] coords = keyToCoords.get(entry.getKey()); | ||
| boolean committed = false; | ||
| openTransaction(); | ||
| try { | ||
| deleteTableColumnStatistics(coords[0], coords[1], coords[2], entry.getValue(), "hive"); | ||
|
Contributor
Author
There was a problem hiding this comment. Same issue as the previous one. Resolved. |
||
| committed = commitTransaction(); | ||
| } finally { | ||
| if (!committed) { | ||
| rollbackTransaction(); | ||
| } | ||
| } | ||
| } | ||
| } | ||
|
|
||
| /** | ||
| * Queries {@code PART_COL_STATS} for rows whose {@code lastAnalyzed} is older than the given | ||
| * threshold. Results are copied into an {@link ArrayList} so they remain accessible after | ||
| * the enclosing JDO transaction is committed. | ||
| * | ||
| * @param pm the JDO persistence manager to use for the query | ||
| * @param lastAnalyzedThreshold epoch seconds; rows with lastAnalyzed below this value are expired | ||
| * @return list of projected rows: [catName, dbName, tblName, partName, colName] | ||
| * @throws Exception if the JDO query fails | ||
| */ | ||
| private List<Object[]> collectExpiredPartitionColStats(PersistenceManager pm, | ||
| long lastAnalyzedThreshold) throws Exception { | ||
| try (Query partQuery = pm.newQuery(MPartitionColumnStatistics.class)) { | ||
| partQuery.setFilter("lastAnalyzed < threshold && engine == \"hive\""); | ||
| partQuery.declareParameters("long threshold"); | ||
| partQuery.setRange(0, 1000); | ||
| partQuery.setResult( | ||
| "partition.table.database.catalogName, " | ||
| + "partition.table.database.name, " | ||
| + "partition.table.tableName, " | ||
| + "partition.partitionName, " | ||
| + "colName, " | ||
| + "partition.table.parameters.get(\"" + STATISTICS_AUTO_DELETION_EXCLUDE_TBLPROPERTY + "\")"); | ||
| @SuppressWarnings("unchecked") | ||
| List<Object[]> rows = (List<Object[]>) partQuery.execute(lastAnalyzedThreshold); | ||
| return new ArrayList<>(rows); | ||
| } | ||
|
Comment on lines
+214
to
+228
Contributor
Author
There was a problem hiding this comment. Same issue as the previous one. Resolved. |
||
| } | ||
|
|
||
| /** | ||
| * Groups expired partition-level column stats by {@code (catName, dbName, tblName, partName)}, | ||
| * then deletes all expired columns for each partition in a single {@link RawStore} call and a | ||
| * dedicated transaction. This avoids duplicate listener events for the same partition. | ||
| * Tables whose exclude property is set to {@code "true"} are skipped entirely. | ||
| * | ||
| * @param expiredPartRows projected rows from {@link #collectExpiredPartitionColStats} | ||
| * @throws Exception if a delete operation fails | ||
| */ | ||
| private void deleteExpiredPartitionColStats(List<Object[]> expiredPartRows) throws Exception { | ||
| // key: catName + SEP + dbName + SEP + tblName + SEP + partName, value: list of expired colNames | ||
| Map<String, List<String>> partToColsMap = new LinkedHashMap<>(); | ||
| Map<String, String[]> keyToCoords = new LinkedHashMap<>(); | ||
| for (Object[] row : expiredPartRows) { | ||
| String catName = (String) row[0]; | ||
| String dbName = (String) row[1]; | ||
| String tblName = (String) row[2]; | ||
| String partName = (String) row[3]; | ||
| String colName = (String) row[4]; | ||
| String excludeVal = (String) row[5]; | ||
| if (Boolean.parseBoolean(excludeVal)) { | ||
| LOG.info("Skipping auto deletion of partition stats for {}.{} due to exclude property.", | ||
| dbName, tblName); | ||
| continue; | ||
| } | ||
| String key = catName + KEY_SEP + dbName + KEY_SEP + tblName + KEY_SEP + partName; | ||
| partToColsMap.computeIfAbsent(key, k -> new ArrayList<>()).add(colName); | ||
| keyToCoords.putIfAbsent(key, new String[]{catName, dbName, tblName, partName}); | ||
| } | ||
|
|
||
| // one transaction per partition, delete all expired columns in a single call | ||
| for (Map.Entry<String, List<String>> entry : partToColsMap.entrySet()) { | ||
| String[] coords = keyToCoords.get(entry.getKey()); | ||
| boolean committed = false; | ||
| openTransaction(); | ||
| try { | ||
| deletePartitionColumnStatistics(coords[0], coords[1], coords[2], | ||
| Collections.singletonList(coords[3]), entry.getValue(), "hive"); | ||
|
Comment on lines
+267
to
+268
Contributor
Author
There was a problem hiding this comment. Same issue as the previous one. Resolved. |
||
| committed = commitTransaction(); | ||
| } finally { | ||
| if (!committed) { | ||
| rollbackTransaction(); | ||
| } | ||
| } | ||
| } | ||
| } | ||
| } | ||
Oops, something went wrong.
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Uh oh!
There was an error while loading. Please reload this page.