[ https://issues.apache.org/jira/browse/HIVE-24275?focusedWorklogId=503205&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-503205 ]
ASF GitHub Bot logged work on HIVE-24275: ----------------------------------------- Author: ASF GitHub Bot Created on: 21/Oct/20 14:04 Start Date: 21/Oct/20 14:04 Worklog Time Spent: 10m Work Description: pvargacl commented on a change in pull request #1583: URL: https://github.com/apache/hive/pull/1583#discussion_r509315841 ########## File path: ql/src/java/org/apache/hadoop/hive/ql/txn/compactor/Cleaner.java ########## @@ -244,7 +249,32 @@ private void removeFiles(String location, ValidWriteIdList writeIdList, Compacti obsoleteDirs.addAll(dir.getAbortedDirectories()); List<Path> filesToDelete = new ArrayList<>(obsoleteDirs.size()); StringBuilder extraDebugInfo = new StringBuilder("["); + boolean delayedCleanupEnabled = conf.getBoolVar(HiveConf.ConfVars.HIVE_COMPACTOR_DELAYED_CLEANUP_ENABLED); + for (Path stat : obsoleteDirs) { + if (delayedCleanupEnabled) { + String filename = stat.toString(); + if (filename.startsWith(AcidUtils.BASE_PREFIX)) { + long writeId = AcidUtils.ParsedBase.parseBase(stat).getWriteId(); + if (ci.type == CompactionType.MINOR) { + LOG.info("Skipping base dir " + stat + " as this cleanup is for minor compaction" + + ", compaction id " + ci.id); + continue; + } else if (writeId > writeIdList.getHighWatermark()) { + LOG.info("Skipping base dir " + stat + " deletion as WriteId of this base dir is" + + " greater than highWaterMark for compaction id " + ci.id); + continue; + } + } + else if (filename.startsWith(AcidUtils.DELTA_PREFIX) || filename.startsWith(AcidUtils.DELETE_DELTA_PREFIX)) { + AcidUtils.ParsedDelta delta = AcidUtils.parsedDelta(stat, fs); Review comment: There is a ParsedDeltaLight in AcidUtils, that is a cheaper way to parse out the maxWriteId. This parsedDelta method will issue a FS call to check if the delta is in raw format ---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: us...@infra.apache.org Issue Time Tracking ------------------- Worklog Id: (was: 503205) Time Spent: 20m (was: 10m) > Configurations to delay the deletion of obsolete files by the Cleaner > --------------------------------------------------------------------- > > Key: HIVE-24275 > URL: https://issues.apache.org/jira/browse/HIVE-24275 > Project: Hive > Issue Type: New Feature > Reporter: Kishen Das > Priority: Major > Labels: pull-request-available > Time Spent: 20m > Remaining Estimate: 0h > > Whenever compaction happens, the cleaner immediately deletes older obsolete > files. In certain cases it would be beneficial to retain these for certain > period. For example : if you are serving the file metadata from cache and > don't want to invalidate the cache during compaction because of performance > reasons. > For this purpose we should introduce a configuration > hive.compactor.delayed.cleanup.enabled, which if enabled will delay the > cleaning up obsolete files. There should be a separate configuration > CLEANER_RETENTION_TIME to specify the duration till which we should retain > these older obsolete files. > It might be beneficial to have one more configuration to decide whether to > retain files involved in an aborted transaction > hive.compactor.aborted.txn.delayed.cleanup.enabled . -- This message was sent by Atlassian Jira (v8.3.4#803005)