Failed to start asynchronous compilation, proceeding with interpreting query

Help Procedural SQL
1

When running a SELECTquery, we got this error:

426714419284 2019-07-15 07:16:02.061   INFO: Query table_name.'SELECT
			column_1, column2
		FROM table_name.column_name
		WHERE
			geography_within_distance(geography_point(@C, @C), location, @C)
			AND indexed_column_1 >= @
		ORDER BY indexed_column_1 DESC
		LIMIT @' submitted for asynchronous compilation

426714423082 2019-07-15 07:16:02.065  ERROR: write() system call (fd=160) failed with errno: 32 (Broken pipe)

426714423128 2019-07-15 07:16:02.065  ERROR: Failed to send file descriptor protocol ack.

426714423154 2019-07-15 07:16:02.065   WARN: Failed to start asynchronous compilation, proceeding with interpreting query: Query  table_name.'SELECT
			column_1, column2
		FROM table_name.column_name
		WHERE
			geography_within_distance(geography_point(@C, @C), location, @C)
			AND indexed_column_1 >= @
		ORDER BY indexed_column_1 DESC
		LIMIT @'

432006499267 2019-07-15 08:44:14.141   INFO: Evicted 0 plans

MemSQL failed to compile the query and not retrying to compile one more time. This made the query is becoming slow.

When we manually run the query one more time using MemSQL Studio, the query is compiled successfully.

What is the root cause of this error? How to proceed from here?

2

Can you share output of following command ?

show variables like ‘%interpreter_mode%’;

Thanks
Jaimin

3 
> show variables like '%interpreter_mode%';
+-------------------------------------------------------------+
| Variable_name                | Value                        |
+-------------------------------------------------------------+
| interpreter_mode             | INTERPRET_FIRST              |
| interpreter_mode_sampling_...| 1000000                      |
+-------------------------------------------------------------+

We are using MemSQL version 6.8.3

4

Did you restart the cluster before you ran the query a second time?

That error looks like the command process (the 2nd memsqld that runs for each node) died. That process is used to run LLVM code-gen. It can get OOM killed by linux sometimes. Do you see any messaged and OOM killer invocations in dmesg?

5

No, we didn’t restart the cluster.

And No, there’s no OOM

6

Can you send us a cluster report to bug-report@memsql.com?