Uploaded image for project: 'Hive'
  1. Hive
  2. HIVE-21154

Investigate using object IDs in Acid HMS schema instead of names

    XMLWordPrintableJSON

Details

    • New Feature
    • Status: Open
    • Major
    • Resolution: Unresolved
    • 1.0.0
    • None
    • Transactions
    • None

    Description

      Currently all Acid related tables in HMS DB (HIVE_LOCKS, TXN_COMPONENTS, etc) use db_name/table_name/partition_name to identify the metastore object that is being tracked (these are potentially long strings, esp partition name. It would improve perf to use object ID such as TBLS.TBL_ID which is exposed in Thrift since HIVE-20556. It would also make handling object rename operations no-op (currently handled in TxnHandler.onRename() from AcidEventListener extends MetaStoreEventListener). This would require significant HMS schema changes and surfacing the ID of Database/Partition objects.

      Need to think how this affects replication.

      Attachments

        Issue Links

          Activity

            People

              rkirtir KIRTI RUGE
              ekoifman Eugene Koifman
              Votes:
              0 Vote for this issue
              Watchers:
              2 Start watching this issue

              Dates

                Created:
                Updated: