[
https://issues.apache.org/jira/browse/DRILL-8479?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17810231#comment-17810231
]
ASF GitHub Bot commented on DRILL-8479:
---------------------------------------
shfshihuafeng opened a new pull request, #2878:
URL: https://github.com/apache/drill/pull/2878
… (#2876)
# [DRILL-8479](https://issues.apache.org/jira/browse/DRILL-8479): mergejoin
leak when Depleting incoming batches throw exception
## Description
when fragment failed, it call close() from MergeJoinBatch. but if
leftIterator.close() throw exception, we could not call rightIterator.close()
to release memory。
## Documentation
(Please describe user-visible changes similar to what should appear in the
Drill documentation.)
## Testing
The test method is the same with link, only one parameter needs to be
modified,
set planner.enable_hashjoin =false to ensure use mergejoin operator
[](https://github.com/apache/drill/pull/2875)
> mergejion memory leak when exception
> -------------------------------------
>
> Key: DRILL-8479
> URL: https://issues.apache.org/jira/browse/DRILL-8479
> Project: Apache Drill
> Issue Type: Bug
> Components: Execution - Relational Operators
> Affects Versions: 1.21.1
> Reporter: shihuafeng
> Priority: Critical
> Attachments: 0001-mergejoin-leak.patch
>
>
> *Describe the bug*
> megerjoin leak when RecordIterator allocate memory exception with
> OutOfMemoryException{*}{*}
> {*}Steps to reproduce the behavior{*}:
> # prepare data for tpch 1s
> # set direct memory 5g
> # set planner.enable_hashjoin =false to ensure use mergejoin operator。
> # set drill.memory.debug.allocator =true (Check for memory leaks )
> # 20 concurrent for tpch sql8
> # when it had OutOfMemoryException or null EXCEPTION , stopped all sql.
> # finding memory leak
> *Expected behavior*
> when all sql sop , we should find direct memory is 0 AND could not
> find leak log like following.
> {code:java}
> Allocator(op:2:0:11:MergeJoinPOP) 1000000/73728/4874240/10000000000
> (res/actual/peak/limit){code}
> *Error detail, log output or screenshots*
> {code:java}
> Unable to allocate buffer of size XX (rounded from XX) due to memory limit
> (). Current allocation: xx{code}
> [^0001-mergejoin-leak.patch]
> sql
> {code:java}
> // code placeholder
> select o_year, sum(case when nation = 'CHINA' then volume else 0 end) /
> sum(volume) as mkt_share from ( select extract(year from o_orderdate) as
> o_year, l_extendedprice * 1.0 as volume, n2.n_name as nation from
> hive.tpch1s.part, hive.tpch1s.supplier, hive.tpch1s.lineitem,
> hive.tpch1s.orders, hive.tpch1s.customer, hive.tpch1s.nation n1,
> hive.tpch1s.nation n2, hive.tpch1s.region where p_partkey = l_partkey and
> s_suppkey = l_suppkey and l_orderkey = o_orderkey and o_custkey = c_custkey
> and c_nationkey = n1.n_nationkey and n1.n_regionkey = r_regionkey and r_name
> = 'ASIA' and s_nationkey = n2.n_nationkey and o_orderdate between date
> '1995-01-01' and date '1996-12-31' and p_type = 'LARGE BRUSHED BRASS') as
> all_nations group by o_year order by o_year
> {code}
>
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
