[ 
https://issues.apache.org/jira/browse/IMPALA-5824?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Tim Armstrong resolved IMPALA-5824.
-----------------------------------
       Resolution: Fixed
    Fix Version/s: Impala 2.10.0

It looks like setting the mem_limit to 150mb (resulting in a buffer pool limit 
of 75mb)  is sufficient. The merge aggregation requires 114mb to run entirely 
in memory so it should reliably spill.

{code}
+---------------------+--------+----------+----------+-------+------------+-----------+---------------+-----------------------+
| Operator            | #Hosts | Avg Time | Max Time | #Rows | Est. #Rows | 
Peak Mem  | Est. Peak Mem | Detail                |
+---------------------+--------+----------+----------+-------+------------+-----------+---------------+-----------------------+
| 05:MERGING-EXCHANGE | 1      | 4.06s    | 4.06s    | 5     | 5          | 0 B 
      | 0 B           | UNPARTITIONED         |
| 02:TOP-N            | 3      | 15.13ms  | 15.30ms  | 15    | 5          | 
76.00 KB  | 200 B         |                       |
| 04:AGGREGATE        | 3      | 2.32s    | 2.34s    | 1.50M | 1.56M      | 
114.16 MB | 34.00 MB      | FINALIZE              |
| 03:EXCHANGE         | 3      | 1.51s    | 1.53s    | 1.50M | 1.56M      | 0 B 
      | 0 B           | HASH(l_orderkey)      |
| 01:AGGREGATE        | 3      | 923.22ms | 1.00s    | 1.50M | 1.56M      | 
114.94 MB | 65.60 MB      | STREAMING             |
| 00:SCAN HDFS        | 3      | 524.84ms | 596.76ms | 6.00M | 6.00M      | 
25.30 MB  | 400.00 MB     | tpch_parquet.lineitem |
+---------------------+--------+----------+----------+-------+------------+-----------+---------------+-----------------------+
{code}

> TestSpilling.test_spilling_aggs failing: SpilledPartitions: 0 (0)
> -----------------------------------------------------------------
>
>                 Key: IMPALA-5824
>                 URL: https://issues.apache.org/jira/browse/IMPALA-5824
>             Project: IMPALA
>          Issue Type: Bug
>          Components: Backend
>    Affects Versions: Impala 2.10.0
>            Reporter: David Knupp
>            Assignee: Tim Armstrong
>            Priority: Critical
>              Labels: broken-build
>             Fix For: Impala 2.10.0
>
>
> Recently seen in ASAN builds. The RUNTIME_PROFILE reports:
> {noformat}
> SpilledPartitions: 0 (0)
> {noformat}
> The test expects a result matching the following regex:
> {noformat}
> 02:35:30  TestSpilling.test_spilling_aggs[exec_option: {'debug_action': None, 
> 'default_spillable_buffer_size': '256k'} | table_format: parquet/none] 
> 02:35:30 [gw3] linux2 -- Python 2.6.6 
> /data/jenkins/workspace/impala-umbrella-build-and-test/repos/Impala/bin/../infra/python/env/bin/python
> 02:35:30 query_test/test_spilling.py:53: in test_spilling_aggs
> 02:35:30     self.run_test_case('QueryTest/spilling-aggs', vector)
> 02:35:30 common/impala_test_suite.py:430: in run_test_case
> 02:35:30     verify_runtime_profile(test_section['RUNTIME_PROFILE'], 
> result.runtime_profile)
> 02:35:30 common/test_result_verifier.py:560: in verify_runtime_profile
> 02:35:30     actual))
> 02:35:30 E   AssertionError: Did not find matches for lines in runtime 
> profile:
> 02:35:30 E   EXPECTED LINES:
> 02:35:30 E   row_regex: .*SpilledPartitions: .* \([1-9][0-9]*\)
> {noformat}
> This has affected these variations of the test.
> {noformat}
> TestSpilling::()::test_spilling_aggs[exec_option: {'debug_action': None, 
> 'default_spillable_buffer_size': '256k'} | table_format: parquet/none]
> TestSpilling::()::test_spilling_aggs[exec_option: {'debug_action': 
> '-1:OPEN:[email protected]', 
> 'default_spillable_buffer_size': '256k'} | table_format: parquet/none]
> TestSpilling::()::test_spilling_aggs[exec_option: {'debug_action': 
> '-1:OPEN:[email protected]', 
> 'default_spillable_buffer_size': '256k'} | table_format: parquet/none]
> {noformat}
> Note that there were other failures in the same test runs similar to 
> IMPALA-5820.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

Reply via email to