kudu-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Casey Ching <ca...@cloudera.com>
Subject Re: Unregister query takes a lot of time
Date Thu, 26 May 2016 03:17:58 GMT
It looks like the latency is coming from a single host

    Fragment F00:
      Instance 834331099f684557:131f56848dd08a9f
(host=node26:22000):(Total: 5s222ms, non-child: 1.996ms, % non-child: 0.04%)

The other fragments are all much faster. I don’t see any clues in the
profile. Maybe something is wrong with that node. Do you see anything
interesting in the Impala or Kudu logs for that host?


On May 25, 2016 at 12:26:57 AM, Darren Hoo (darren.hoo@gmail.com) wrote:

Doing a simple primary key lookup, Unregister query takes 5 seconds,  it
seems too long to me.
what is it doing?


Operator       #Hosts   Avg Time  Max Time  #Rows  Est. #Rows   Peak Mem
 Est. Peak Mem  Detail
-------------------------------------------------------------------------------------------------------
01:EXCHANGE         1   19.852us  19.852us      1           1          0
     -1.00 B  UNPARTITIONED
00:SCAN KUDU       25  259.654ms   5s220ms      1           1  104.00 KB
           0
    Planner Timeline: 35.247ms
       - Analysis finished: 601.876us (601.876us)
       - Equivalence classes computed: 789.600us (187.724us)
       - Single node plan created: 34.156ms (33.367ms)
       - Runtime filters computed: 34.167ms (11.068us)
       - Distributed plan created: 34.697ms (529.183us)
       - Planning finished: 35.247ms (550.736us)
    Query Timeline: 5s286ms
       - Start execution: 236.944us (236.944us)
       - Planning finished: 37.069ms (36.833ms)
       - Ready to start 25 remote fragments: 43.452ms (6.382ms)
       - All 25 remote fragments started: 48.160ms (4.707ms)
       - Rows available: 56.501ms (8.340ms)
       - First row fetched: 151.018ms (94.517ms)
       - Unregister query: 5s275ms (5s124ms)
  ImpalaServer:


full profile attached

Mime
View raw message