Uploaded image for project: 'IMPALA'
  2. IMPALA-4559

Impala query performance issues



    • Type: Task
    • Status: Resolved
    • Priority: Minor
    • Resolution: Information Provided
    • Affects Version/s: Impala 2.3.0
    • Fix Version/s: None
    • Component/s: Perf Investigation
    • Labels:


      I am running a Query which returns 5 rows
      select distinct date_key from tbl_date limit 5; /the table has a few hundred rows with 1 partition/

      The Query info is

      Planning Wait Time: 18.8m Planning Wait Time Percentage: 100

      Query Timeline Start execution: 95.55us (95.55us)
      Planning finished: 18.8m (18.8m)
      Submit for admission: 18.8m (345.26us)
      Completed admission: 18.8m (107.67us)
      Ready to start remote fragments: 18.8m (1ms)
      Remote fragments started: 18.8m (1.13s)
      Rows available: 18.9m (4.15s)
      First row fetched: 18.9m (7ms)
      Unregister query: 18.9m (33ms)

      The issue is why is the "Planning Wait Time" so high ?
      planning wait time is defined as - The total amount of time the query spent waiting for planning to complete.

      what I don't understand with such a simple query why is spending so much time planning ?
      also a lot of queries which are slow have high Planning Wait Time Percentage why ?
      can I get some more info on Planning Wait Time metric?

      Query Info -
      Start Time: Nov 30, 2016 11:17:13 AM
      End Time: Nov 30, 2016 11:36:05 AM
      Duration: 18m, 52s
      Rows Produced: 5
      Aggregate Peak Memory Usage: 125.2 MiB
      Bytes Streamed: 17.0 KiB
      Client Fetch Wait Time: 8ms
      Client Fetch Wait Time Percentage: 0
      Estimated per Node Peak Memory: 256.0 MiB
      HDFS Average Scan Range: 3.5 KiB
      HDFS Bytes Read: 304.9 KiB
      HDFS Bytes Read From Cache: 0 B
      HDFS Bytes Read From Cache Percentage: 0
      HDFS Local Bytes Read: 304.9 KiB
      HDFS Local Bytes Read Percentage: 100
      HDFS Remote Bytes Read: 0 B
      HDFS Remote Bytes Read Percentage: 0
      HDFS Scanner Average Read Throughput: 84.2 MiB/s
      HDFS Short Circuit Bytes Read: 304.9 KiB
      HDFS Short Circuit Bytes Read Percentage: 100
      Impala Version: impalad version 2.3.0-cdh5.5.2 RELEASE (build cc1125f10419a7269366f7f950f57b24b07acd64)
      Memory Accrual: 612,880,244 byte seconds
      Per Node Peak Memory Usage: 7.5 MiB
      Planning Wait Time: 18.8m
      Planning Wait Time Percentage: 100
      Pool: root.default
      Query Status: OK
      Session ID: 83470f9951838902:61749525b39b70b7
      Session Type: HIVESERVER2
      Statistics Missing: true
      Threads: CPU Time: 1.4m
      Threads: CPU Time Percentage: 31
      Threads: Network Receive Wait Time: 2.8m
      Threads: Network Receive Wait Time Percentage: 64
      Threads: Network Send Wait Time: 1.16s
      Threads: Network Send Wait Time Percentage: 0
      Threads: Storage Wait Time: 12.72s
      Threads: Storage Wait Time Percentage: 5
      Threads: Total Time: 4.4m
      Work CPU Time: 1.4m




            • Assignee:
              araw012_impala_7e72 anil
            • Votes:
              0 Vote for this issue
              4 Start watching this issue


              • Created: