[DRILL-3665] Deadlock while executing CTAS that runs out of memory

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Duplicate
• Affects Version/s: 1.2.0
• Fix Version/s: 1.11.0
• Component/s: Execution - Flow
• Assignee: Roman Kulyk
• Reporter: Victoria Markman
• Created: 18/Aug/15 17:58
• Updated: 03/Aug/17 17:57
• Resolved: 03/Aug/17 17:57

Description

I had a query running out of memory during CTAS and after that drillbit was rendered unusable:

0: jdbc:drill:schema=dfs> create table lineitem as select
. . . . . . . . . . . . >     cast(columns[0] as int) l_orderkey,
. . . . . . . . . . . . >     cast(columns[1] as int) l_partkey,
. . . . . . . . . . . . >     cast(columns[2] as int) l_suppkey,
. . . . . . . . . . . . >     cast(columns[3] as int) l_linenumber,
. . . . . . . . . . . . >     cast(columns[4] as double) l_quantity,
. . . . . . . . . . . . >     cast(columns[5] as double) l_extendedprice,
. . . . . . . . . . . . >     cast(columns[6] as double) l_discount,
. . . . . . . . . . . . >     cast(columns[7] as double) l_tax,
. . . . . . . . . . . . >     cast(columns[8] as varchar(200)) l_returnflag,
. . . . . . . . . . . . >     cast(columns[9] as varchar(200)) l_linestatus,
. . . . . . . . . . . . >     cast(columns[10] as date) l_shipdate,
. . . . . . . . . . . . >     cast(columns[11] as date) l_commitdate,
. . . . . . . . . . . . >     cast(columns[12] as date) l_receiptdate,
. . . . . . . . . . . . >     cast(columns[13] as varchar(200)) l_shipinstruct,
. . . . . . . . . . . . >     cast(columns[14] as varchar(200)) l_shipmode,
. . . . . . . . . . . . >     cast(columns[15] as varchar(200)) l_comment
. . . . . . . . . . . . > from `lineitem.dat`;
Error: RESOURCE ERROR: One or more nodes ran out of memory while executing the query.
Fragment 1:10
[Error Id: 11084315-5388-4500-b165-642a5f595ebf on atsqa4-133.qa.lab:31010] (state=,code=0)

Here is drill's behavior after that:

1. Tried to run: "select * from sys.options" in the same sqlline session - hangs.

2. Was able to start sqlline and connect to drillbit:

• If you try running anything on this connection: it hangs.
• Issue ^C --> you will get result if you are lucky (these queries will appear as: "CANCELLATION_REQUESTED" on WebUI)
  (I only tried querying sys.memory, sys.options which possibly have a different code path than queries from actual user data)
• If you are not lucky, you will get this error below:

          0: jdbc:drill:schema=dfs> show files;
          java.lang.RuntimeException: java.sql.SQLException: Unexpected RuntimeException: java.lang.IllegalArgumentException: Buffer has negative reference count.
          at sqlline.IncrementalRows.hasNext(IncrementalRows.java:73)
          at sqlline.TableOutputFormat$ResizingRowsProvider.next(TableOutputFormat.java:87)
          at sqlline.TableOutputFormat.print(TableOutputFormat.java:118)
          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)
  

or maybe something like this:

0: jdbc:drill:schema=dfs> select count(*) from nation group by n_regionkey;
Error: CONNECTION ERROR: Exceeded timeout (5000) while waiting send intermediate work fragments to remote nodes. Sent 1 and only heard response back from 0 nodes.
[Error Id: 6abce8e9-78a1-4b3d-bcec-503930482b40 on atsqa4-133.qa.lab:31010] (state=,code=0)

I'm attaching results of a jstack and drillbit.log and so far I was not able to reproduce this problem again (working on it).

Issue Links

• duplicates: DRILL-5599 Notify StatusHandlerListener that batch sending has failed even if channel is still open