[DRILL-3265] Query with window function and sort below that spills to disk runs out of memory

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.0.0
• Fix Version/s: 1.1.0
• Component/s: Execution - Relational Operators
• Labels: window_function
• Assignee: Abdel Hakim Deneche
• Reporter: Victoria Markman
• Created: 09/Jun/15 17:46
• Updated: 23/Jun/15 00:02
• Resolved: 23/Jun/15 00:02

Description

Failure:

0: jdbc:drill:schema=dfs> select
. . . . . . . . . . . . >   sum(ss_quantity) over(partition by ss_store_sk order by ss_sold_date_sk)
. . . . . . . . . . . . > from store_sales;
java.lang.RuntimeException: java.sql.SQLException: RESOURCE ERROR: One or more nodes ran out of memory while executing the query.

Fragment 1:13

[Error Id: 72609220-e431-41fc-8505-8f9740c96153 on atsqa4-133.qa.lab:31010]
        at sqlline.IncrementalRows.hasNext(IncrementalRows.java:73)
        at sqlline.TableOutputFormat$ResizingRowsProvider.next(TableOutputFormat.java:85)
        at sqlline.TableOutputFormat.print(TableOutputFormat.java:116)
        at sqlline.SqlLine.print(SqlLine.java:1583)
        at sqlline.Commands.execute(Commands.java:852)
        at sqlline.Commands.sql(Commands.java:751)
        at sqlline.SqlLine.dispatch(SqlLine.java:738)
        at sqlline.SqlLine.begin(SqlLine.java:612)
        at sqlline.SqlLine.start(SqlLine.java:366)
        at sqlline.SqlLine.main(SqlLine.java:259)

Failure looks legitimate from customer point of view (granted that we know that we can run out of memory in some cases)

However, there are couple of problems with this scenario:

1. It looks like we are running out of memory during disk based sort

2015-06-09 16:55:45,693 [2a88e633-b714-49a1-c36a-509a9c817c77:frag:1:19] WARN  o.a.d.e.p.i.xsort.ExternalSortBatch - Starting to merge. 3311 batch groups. Current allocated memory: 17456644
2015-06-09 16:55:45,763 [2a88e633-b714-49a1-c36a-509a9c817c77:frag:1:13] WARN  o.a.d.exec.memory.BufferAllocator - Unable to allocate buffer of size 20833 due to memory limit. Current allocation: 19989451
java.lang.Exception: null
        at org.apache.drill.exec.memory.TopLevelAllocator$ChildAllocator.buffer(TopLevelAllocator.java:253) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.memory.TopLevelAllocator$ChildAllocator.buffer(TopLevelAllocator.java:273) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.vector.UInt1Vector.allocateNew(UInt1Vector.java:171) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.vector.NullableIntVector.allocateNew(NullableIntVector.java:204) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.vector.AllocationHelper.allocateNew(AllocationHelper.java:56) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.test.generated.PriorityQueueCopierGen139.allocateVectors(PriorityQueueCopierTemplate.java:123) [na:na]
        at org.apache.drill.exec.test.generated.PriorityQueueCopierGen139.next(PriorityQueueCopierTemplate.java:66) [na:na]
        at org.apache.drill.exec.physical.impl.xsort.ExternalSortBatch.innerNext(ExternalSortBatch.java:224) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.record.AbstractRecordBatch.next(AbstractRecordBatch.java:146) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.record.AbstractRecordBatch.next(AbstractRecordBatch.java:105) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.record.AbstractRecordBatch.next(AbstractRecordBatch.java:95) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.record.AbstractSingleRecordBatch.innerNext(AbstractSingleRecordBatch.java:51) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.physical.impl.svremover.RemovingRecordBatch.innerNext(RemovingRecordBatch.java:92) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.record.AbstractRecordBatch.next(AbstractRecordBatch.java:146) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.record.AbstractRecordBatch.next(AbstractRecordBatch.java:105) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.record.AbstractRecordBatch.next(AbstractRecordBatch.java:95) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.physical.impl.window.WindowFrameRecordBatch.innerNext(WindowFrameRecordBatch.java:123) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.record.AbstractRecordBatch.next(AbstractRecordBatch.java:146) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.record.AbstractRecordBatch.next(AbstractRecordBatch.java:105) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.record.AbstractRecordBatch.next(AbstractRecordBatch.java:95) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.record.AbstractSingleRecordBatch.innerNext(AbstractSingleRecordBatch.java:51) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.physical.impl.project.ProjectRecordBatch.innerNext(ProjectRecordBatch.java:129) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.record.AbstractRecordBatch.next(AbstractRecordBatch.java:146) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.physical.impl.BaseRootExec.next(BaseRootExec.java:83) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.physical.impl.SingleSenderCreator$SingleSenderRootExec.innerNext(SingleSenderCreator.java:95) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.physical.impl.BaseRootExec.next(BaseRootExec.java:73) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.work.fragment.FragmentExecutor$1.run(FragmentExecutor.java:259) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.exec.work.fragment.FragmentExecutor$1.run(FragmentExecutor.java:253) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at java.security.AccessController.doPrivileged(Native Method) [na:1.7.0_71]
        at javax.security.auth.Subject.doAs(Subject.java:415) [na:1.7.0_71]
        at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1566) [hadoop-common-2.5.1-mapr-1503.jar:na]
        at org.apache.drill.exec.work.fragment.FragmentExecutor.run(FragmentExecutor.java:253) [drill-java-exec-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at org.apache.drill.common.SelfCleaningRunnable.run(SelfCleaningRunnable.java:38) [drill-common-1.1.0-SNAPSHOT-rebuffed.jar:1.1.0-SNAPSHOT]
        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145) [na:1.7.0_71]
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615) [na:1.7.0_71]
        at java.lang.Thread.run(Thread.java:745) [na:1.7.0_71]
2015-06-09 16:55:45,765 [2a88e633-b714-49a1-c36a-509a9c817c77:frag:1:13] INFO  o.a.d.c.exceptions.UserException - User Error Occurred
org.apache.drill.common.exceptions.UserException: RESOURCE ERROR: One or more nodes ran out of memory while executing the query.

2. The same sort running by itself and spilling to disk finishes successfully.

3. There is an indication in drillbit.log that memory leak has occurred.

2015-06-09 16:55:46,053 [2a88e633-b714-49a1-c36a-509a9c817c77:frag:1:11] ERROR o.a.d.c.exceptions.UserException - SYSTEM ERROR: java.lang.IllegalStateException: Failure while closing accountor.  Expected private and shared pools to be set to initial values.  However, one or more were not.  Stats are
        zone    init    allocated       delta
        private 10000000        10636   9989364
        shared  10000000        3114634 6885366.

Fragment 1:11

[Error Id: 607d4f33-dda8-4cd9-8489-e34e914737e9 on atsqa4-133.qa.lab:31010]
org.apache.drill.common.exceptions.UserException: SYSTEM ERROR: java.lang.IllegalStateException: Failure while closing accountor.  Expected private and shared pools to be set to initial values.  However, one or more were not.  Stats are
        zone    init    allocated       delta
        private 10000000        10636   9989364
        shared  10000000        3114634 6885366.

Fragment 1:11

3. Ocasionally, after executing the same query after it failed, I see an error on failing to create a spill directory even though I have enough disk space in my spill directory that is configured to be on a local file system. (remember, sort by itself runs fine)

0: jdbc:drill:schema=dfs> select 
. . . . . . . . . . . . >   sum(ss_quantity) over(partition by ss_store_sk order by ss_sold_date_sk) 
. . . . . . . . . . . . > from store_sales;
java.lang.RuntimeException: java.sql.SQLException: SYSTEM ERROR: java.io.IOException: Mkdirs failed to create /tmp/drill/spill/2a88e74d-df47-146d-0cf9-8773303a7f37/major_fragment_1/minor_fragment_0/operator_4 (exists=false, cwd=file:///opt/mapr/drill/drill-1.1.0/bin)

Additional information:
Plan for the query that runs out of memory:

0: jdbc:drill:schema=dfs> explain plan for select
. . . . . . . . . . . . >   sum(ss_quantity) over(partition by ss_store_sk order by ss_sold_date_sk)
. . . . . . . . . . . . > from store_sales;
+------+------+
| text | json |
+------+------+
| 00-00    Screen
00-01      UnionExchange
01-01        Project(EXPR$0=[CASE(>($3, 0), CAST($4):ANY, null)])
01-02          Window(window#0=[window(partition {1} order by [2] range between UNBOUNDED PRECEDING and CURRENT ROW aggs [COUNT($0), $SUM0($0)])])01-03            SelectionVectorRemover
01-04              Sort(sort0=[$1], sort1=[$2], dir0=[ASC], dir1=[ASC])
01-05                Project(ss_quantity=[$0], ss_store_sk=[$1], ss_sold_date_sk=[$2])
01-06                  HashToRandomExchange(dist0=[[$1]])
02-01                    UnorderedMuxExchange
03-01                      Project(ss_quantity=[$0], ss_store_sk=[$1], ss_sold_date_sk=[$2], E_X_P_R_H_A_S_H_F_I_E_L_D=[castInt(hash64AsDouble($1))])
03-02                        Project(ss_quantity=[$2], ss_store_sk=[$1], ss_sold_date_sk=[$0])
03-03                          Scan(groupscan=[ParquetGroupScan [entries=[ReadEntryWithPath [path=maprfs:///tpcds100/parquet/store_sales]], selectionRoot=/tpcds100/parquet/store_sales, numFiles=1, columns=[`ss_qua
ntity`, `ss_store_sk`, `ss_sold_date_sk`]]])

The same sort executes succesfully when window function is not present:

0: jdbc:drill:schema=dfs> explain plan for select ss_quantity, ss_store_sk, ss_sold_date_sk from store_sales order by ss_store_sk, ss_sold_date_sk;
+------+------+
| text | json |
+------+------+
| 00-00    Screen
00-01      Project(ss_quantity=[$0], ss_store_sk=[$1], ss_sold_date_sk=[$2])
00-02        SingleMergeExchange(sort0=[1 ASC], sort1=[2 ASC])
01-01          SelectionVectorRemover
01-02            Sort(sort0=[$1], sort1=[$2], dir0=[ASC], dir1=[ASC])
01-03              Project(ss_quantity=[$0], ss_store_sk=[$1], ss_sold_date_sk=[$2])
01-04                HashToRandomExchange(dist0=[[$1]], dist1=[[$2]])
02-01                  UnorderedMuxExchange
03-01                    Project(ss_quantity=[$0], ss_store_sk=[$1], ss_sold_date_sk=[$2], E_X_P_R_H_A_S_H_F_I_E_L_D=[castInt(hash64AsDouble($2, hash64AsDouble($1)))])
03-02                      Project(ss_quantity=[$2], ss_store_sk=[$1], ss_sold_date_sk=[$0])
03-03                        Scan(groupscan=[ParquetGroupScan [entries=[ReadEntryWithPath [path=maprfs:///tpcds100/parquet/store_sales]], selectionRoot=/tpcds100/parquet/store_sales, numFiles=1, columns=[`ss_quantity`, `ss_store_sk`, `ss_sold_date_sk`]]])

To reproduce:

1. Single drillbit
2. Configuration:
DRILL_MAX_DIRECT_MEMORY="8G"
DRILL_HEAP="4G"
3. Dataset: tpcds100 parquet
4. Run query:

select 
  sum(ss_quantity) over(partition by ss_store_sk order by ss_sold_date_sk) 
from store_sales;