[ https://issues.apache.org/jira/browse/PHOENIX-6965?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Istvan Toth updated PHOENIX-6965: --------------------------------- Description: UngroupedAggregateRegionScanner.insertEmptyKeyValue() is supposed to generate one empty cell for each timestmap. However, it never adds the generated timestamp to the processed map, and generates an empty cell for each existing cell. IUC, these are all sent to the the RS and written to WAL, and are only consolidated in the memstore, leading to a lot of wasted cpu, memory, io, and network resources. was: UngroupedAggregateRegionScanner.insertEmptyKeyValue() is supposed to generate one empty cell for each timestmap. However, it never adds the generated timestamp to the controll map, and generated an empty cell for each existing cell. IIRC, these are all set to the the RS an d to WAL, and are only consolidated in the memstore, leading to a lot of wasted cpu, memory, io, and network resources. > UngroupedAggregateRegionScanner.insertEmptyKeyValue() generates too many cells > ------------------------------------------------------------------------------ > > Key: PHOENIX-6965 > URL: https://issues.apache.org/jira/browse/PHOENIX-6965 > Project: Phoenix > Issue Type: Bug > Components: core > Affects Versions: 5.2.0, 5.1.3 > Reporter: Istvan Toth > Assignee: Istvan Toth > Priority: Critical > > UngroupedAggregateRegionScanner.insertEmptyKeyValue() is supposed to generate > one empty cell for each timestmap. > However, it never adds the generated timestamp to the processed map, and > generates an empty cell for each existing cell. > IUC, these are all sent to the the RS and written to WAL, and are only > consolidated in the memstore, leading to a lot of wasted cpu, memory, io, and > network resources. -- This message was sent by Atlassian Jira (v8.20.10#820010)