[
https://issues.apache.org/jira/browse/HIVE-26788?focusedWorklogId=832984&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-832984
]
ASF GitHub Bot logged work on HIVE-26788:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 13/Dec/22 08:53
Start Date: 13/Dec/22 08:53
Worklog Time Spent: 10m
Work Description: deniskuzZ commented on code in PR #3812:
URL: https://github.com/apache/hive/pull/3812#discussion_r1046820463
##########
ql/src/java/org/apache/hadoop/hive/ql/txn/compactor/Worker.java:
##########
@@ -428,6 +430,11 @@ protected Boolean findNextCompactionAndExecute(boolean
collectGenericStats, bool
AcidMetricService.updateMetricsFromWorker(ci.dbname, ci.tableName,
ci.partName, ci.type,
dir.getCurrentDirectories().size(), dir.getDeleteDeltas().size(),
conf, msc);
+
+ // Get columns with stats at the end.
+ // This will save computations if compaction has failed.
+ columnList =
msc.findColumnsWithStats(CompactionInfo.compactionInfoToStruct(ci));
Review Comment:
Why don't we fetch the column list right before we call
`statsUpdater.gatherStats`? Also, it is suboptimal to call it always as it's
not cheap, only when auto-stats is disabled and it's a MAJOR compaction. Maybe
would be even better to do it in StatsUpdater
Issue Time Tracking
-------------------
Worklog Id: (was: 832984)
Time Spent: 2h 50m (was: 2h 40m)
> Update stats of table/partition using noscan operation when stats autogather
> is enabled
> ---------------------------------------------------------------------------------------
>
> Key: HIVE-26788
> URL: https://issues.apache.org/jira/browse/HIVE-26788
> Project: Hive
> Issue Type: Improvement
> Reporter: Sourabh Badhya
> Assignee: Sourabh Badhya
> Priority: Major
> Labels: pull-request-available
> Time Spent: 2h 50m
> Remaining Estimate: 0h
>
> Currently, statistics are always updated for major compaction and never
> updated for minor compaction since minor compaction performs little updates
> on the statistics (such as number of files in table/partition & total size of
> the table/partition). It is better to utilize NOSCAN operation for both
> these compaction since NOSCAN operations performs faster update of statistics
> and updates the relevant fields such as number of files & total sizes of the
> table/partitions if stats autogather is enabled. If autostats gather is
> disabled then go for full update of statistics.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
