[jira] [Created] (FLINK-19103) The PushPartitionIntoTableSourceScanRule will lead a performance problem when there are still many partitions after pruning

[Shang Yuanchun (Jira)]

fa zheng created FLINK-19103:
--------------------------------

             Summary: The PushPartitionIntoTableSourceScanRule will lead a performance problem when there are still many partitions after pruning
                 Key: FLINK-19103
                 URL: https://issues.apache.org/jira/browse/FLINK-19103
             Project: Flink
          Issue Type: Improvement
          Components: Table SQL / Planner
    Affects Versions: 1.11.1, 1.10.2
            Reporter: fa zheng
             Fix For: 1.12.0


The PushPartitionIntoTableSourceScanRule will obtain new statistic after pruning, however, it use a for loop to get statistics of each partitions and then merge them together. During this process, flink will try to call metastore's interface four times in one loop. When remaining partitions are huge, it spends a lot of time to get new statistic.
 

{code:scala}
    val newStatistic = {
      val tableStats = catalogOption match {
        case Some(catalog) =>
          def mergePartitionStats(): TableStats = {
            var stats: TableStats = null
            for (p <- remainingPartitions) {
              getPartitionStats(catalog, tableIdentifier, p) match {
                case Some(currStats) =>
                  if (stats == null) {
                    stats = currStats
                  } else {
                    stats = stats.merge(currStats)
                  }
                case None => return null
              }
            }
            stats
          }
          mergePartitionStats()
        case None => null
      }
      FlinkStatistic.builder().statistic(statistic).tableStats(tableStats).build()
    }
{code}




--
This message was sent by Atlassian Jira
(v8.3.4#803005)