hive-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Gopal V (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HIVE-21818) CBO: Copying TableRelOptHiveTable has metastore traffic
Date Sat, 01 Jun 2019 01:11:00 GMT

     [ https://issues.apache.org/jira/browse/HIVE-21818?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Gopal V updated HIVE-21818:
---------------------------
    Description: 
While the optimizer is running, whenever it makes a copy of the TableScan to perform potential
rewrites, there is Hive metastore traffic from inside CBO optimizer.

{code}
  public RelOptHiveTable(RelOptSchema calciteSchema, RelDataTypeFactory typeFactory, List<String>
qualifiedTblName,
      RelDataType rowType, Table hiveTblMetadata, List<ColumnInfo> hiveNonPartitionCols,
      List<ColumnInfo> hivePartitionCols, List<VirtualColumn> hiveVirtualCols,
HiveConf hconf,
      Map<String, PrunedPartitionList> partitionCache, Map<String, ColumnStatsList>
colStatsCache,
      AtomicInteger noColsMissingStats) { ....


    Pair<List<ImmutableBitSet>, List<ImmutableBitSet>> constraintKeys =
generateKeys();
    this.keys = constraintKeys.left;
    this.nonNullablekeys = constraintKeys.right;
    this.referentialConstraints = generateReferentialConstraints();
}
{code}

This is triggered from the rules as the partition pruner

{code}
  protected void perform(RelOptRuleCall call, Filter filter,
      HiveTableScan tScan) {
    // Original table
    RelOptHiveTable hiveTable = (RelOptHiveTable) tScan.getTable();

    // Copy original table scan and table
    HiveTableScan tScanCopy = tScan.copyIncludingTable(tScan.getRowType());
    RelOptHiveTable hiveTableCopy = (RelOptHiveTable) tScanCopy.getTable();
...

    if (StringUtils.equals(hiveTableCopy.getPartitionListKey(), hiveTable.getPartitionListKey()))
{
      // Nothing changed, we do not need to produce a new expression
      return;
    }

{code}

  was:
While the optimizer is running, whenever it makes a copy of the TableScan to perform potential
rewrites, there is Hive metastore traffic from inside CBO optimizer.

{code}
  public RelOptHiveTable(RelOptSchema calciteSchema, RelDataTypeFactory typeFactory, List<String>
qualifiedTblName,
      RelDataType rowType, Table hiveTblMetadata, List<ColumnInfo> hiveNonPartitionCols,
      List<ColumnInfo> hivePartitionCols, List<VirtualColumn> hiveVirtualCols,
HiveConf hconf,
      Map<String, PrunedPartitionList> partitionCache, Map<String, ColumnStatsList>
colStatsCache,
      AtomicInteger noColsMissingStats) { ....


    Pair<List<ImmutableBitSet>, List<ImmutableBitSet>> constraintKeys =
generateKeys();
    this.keys = constraintKeys.left;
    this.nonNullablekeys = constraintKeys.right;
    this.referentialConstraints = generateReferentialConstraints();
}
{code}

This is triggered from the rules as the partition pruner

{code}
  protected void perform(RelOptRuleCall call, Filter filter,
      HiveTableScan tScan) {
    // Original table
    RelOptHiveTable hiveTable = (RelOptHiveTable) tScan.getTable();

    // Copy original table scan and table
    HiveTableScan tScanCopy = tScan.copyIncludingTable(tScan.getRowType());
    RelOptHiveTable hiveTableCopy = (RelOptHiveTable) tScanCopy.getTable();
{code}


> CBO: Copying TableRelOptHiveTable has metastore traffic
> -------------------------------------------------------
>
>                 Key: HIVE-21818
>                 URL: https://issues.apache.org/jira/browse/HIVE-21818
>             Project: Hive
>          Issue Type: Bug
>          Components: CBO
>            Reporter: Gopal V
>            Assignee: Gopal V
>            Priority: Major
>
> While the optimizer is running, whenever it makes a copy of the TableScan to perform
potential rewrites, there is Hive metastore traffic from inside CBO optimizer.
> {code}
>   public RelOptHiveTable(RelOptSchema calciteSchema, RelDataTypeFactory typeFactory,
List<String> qualifiedTblName,
>       RelDataType rowType, Table hiveTblMetadata, List<ColumnInfo> hiveNonPartitionCols,
>       List<ColumnInfo> hivePartitionCols, List<VirtualColumn> hiveVirtualCols,
HiveConf hconf,
>       Map<String, PrunedPartitionList> partitionCache, Map<String, ColumnStatsList>
colStatsCache,
>       AtomicInteger noColsMissingStats) { ....
>     Pair<List<ImmutableBitSet>, List<ImmutableBitSet>> constraintKeys
= generateKeys();
>     this.keys = constraintKeys.left;
>     this.nonNullablekeys = constraintKeys.right;
>     this.referentialConstraints = generateReferentialConstraints();
> }
> {code}
> This is triggered from the rules as the partition pruner
> {code}
>   protected void perform(RelOptRuleCall call, Filter filter,
>       HiveTableScan tScan) {
>     // Original table
>     RelOptHiveTable hiveTable = (RelOptHiveTable) tScan.getTable();
>     // Copy original table scan and table
>     HiveTableScan tScanCopy = tScan.copyIncludingTable(tScan.getRowType());
>     RelOptHiveTable hiveTableCopy = (RelOptHiveTable) tScanCopy.getTable();
> ...
>     if (StringUtils.equals(hiveTableCopy.getPartitionListKey(), hiveTable.getPartitionListKey()))
{
>       // Nothing changed, we do not need to produce a new expression
>       return;
>     }
> {code}



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Mime
View raw message