[
https://issues.apache.org/jira/browse/DRILL-6543?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16531798#comment-16531798
]
ASF GitHub Bot commented on DRILL-6543:
---------------------------------------
ilooner commented on a change in pull request #1351: DRILL-6543: Disable Hash
Join fallback, add percent_reserved_allowance_from_direct
URL: https://github.com/apache/drill/pull/1351#discussion_r199909294
##########
File path:
exec/java-exec/src/main/java/org/apache/drill/exec/ExecConstants.java
##########
@@ -490,6 +491,31 @@ private ExecConstants() {
public static DoubleValidator PERCENT_MEMORY_PER_QUERY = new
RangeDoubleValidator(
PERCENT_MEMORY_PER_QUERY_KEY, 0, 1.0);
+ /**
+ * Enforce reserving some percentage of the JVM's Direct Memory for the
+ * non-buffered operators. I.e., whether using max_query_memory_per_node or
+ * percent_per_query, the memory result (for all the buffered operators) can
+ * not exceed the size of the Direct Memory minus this reserved allowance.
+ * <p>
+ * This allowance is needed to prevent a potential OOM. In case the total
memory
+ * promised for the buffered operators is very close to the Direct Memory's
size,
+ * then if some non-buffered operators (e.g., scanners) also grab
significant memory,
+ * then the remaining memory is less than the promised memory, leading to an
OOM.
+ * </p>
+ * <p>
+ * Note that this enforcement is only good for the non concurrent case. For
multiple
+ * concurrent queries, other queries may grab some unaccounted for Direct
Memory. For
+ * concurrent work, better set the option "planner.memory.percent_per_query"
correctly.
+ * </p>
+ * <p>
+ * DEFAULT: 25%
+ * </p>
+ */
Review comment:
Thanks for adding this documentation.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
> Option for memory mgmt: Reserve allowance for non-buffered
> ----------------------------------------------------------
>
> Key: DRILL-6543
> URL: https://issues.apache.org/jira/browse/DRILL-6543
> Project: Apache Drill
> Issue Type: Improvement
> Components: Execution - Relational Operators
> Affects Versions: 1.13.0
> Reporter: Boaz Ben-Zvi
> Assignee: Boaz Ben-Zvi
> Priority: Major
> Fix For: 1.15.0
>
>
> Introduce a new option to enforce/remind users to reserve some allowance when
> budgeting their memory:
> The problem: When the "planner.memory.max_query_memory_per_node" (MQMPN)
> option is set equal (or "nearly equal") to the allocated *Direct Memory*, an
> OOM is still possible. The reason is that the memory used by the
> "non-buffered" operators is not taken into account.
> For example, MQMPN == Direct-Memory == 100 MB. Run a query with 5 buffered
> operators (e.g., 5 instances of a Hash-Join), so each gets "promised" 20 MB.
> When other non-buffered operators (e.g., a Scanner, or a Sender) also grab
> some of the Direct Memory, then less than 100 MB is left available. And if
> all those 5 Hash-Joins are pushing their limits, then one HJ may have only
> allocated 12MB so far, but on the next 1MB allocation it will hit an OOM
> (from the JVM, as all the 100MB Direct memory is already used).
> A solution -- a new option to _*reserve*_ some of the Direct Memory for those
> non-buffered operators (e.g., default %25). This *allowance* may prevent many
> of the cases like the example above. The new option would return an error
> (when a query initiates) if the MQMPN is set too high. Note that this option
> +can not+ address concurrent queries.
> This should also apply to the alternative for the MQMPN - the
> {{"planner.memory.percent_per_query"}} option (PPQ). The PPQ does not
> _*reserve*_ such memory (e.g., can set it to %100); only its documentation
> clearly explains this issue (that doc suggests reserving %50 allowance, as it
> was written when the Hash-Join was non-buffered; i.e., before spill was
> implemented).
> The memory given to the buffered operators is the highest calculated between
> the MQMPN and the PPQ. The new reserve option would verify that this figure
> allows the allowance.
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
