[
https://issues.apache.org/jira/browse/DRILL-5457?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16027244#comment-16027244
]
ASF GitHub Bot commented on DRILL-5457:
---------------------------------------
Github user paul-rogers commented on a diff in the pull request:
https://github.com/apache/drill/pull/822#discussion_r118813857
--- Diff:
exec/java-exec/src/main/java/org/apache/drill/exec/physical/impl/aggregate/HashAggTemplate.java
---
@@ -400,114 +782,411 @@ public IterOutcome getOutcome() {
@Override
public int getOutputCount() {
- // return outputCount;
return lastBatchOutputCount;
}
@Override
public void cleanup() {
- if (htable != null) {
- htable.clear();
- htable = null;
- }
+ if ( schema == null ) { return; } // not set up; nothing to clean
+ for ( int i = 0; i < numPartitions; i++) {
+ if (htables[i] != null) {
+ htables[i].clear();
+ htables[i] = null;
+ }
+ if ( batchHolders[i] != null) {
+ for (BatchHolder bh : batchHolders[i]) {
+ bh.clear();
+ }
+ batchHolders[i].clear();
+ batchHolders[i] = null;
+ }
+
+ // delete any (still active) output spill file
+ if ( outputStream[i] != null && spillFiles[i] != null) {
+ try {
+ spillSet.delete(spillFiles[i]);
--- End diff --
This code makes me wonder... How many output streams are open at any one
time? Drill is a highly concurrent system: we could have 1000s of fragments. If
each has, say, a hash agg with 16 partitions, do we run the risk of 16,000 open
file handles? Or, is the file handle opened only when needed for reading or
writing?
In general, as we add more spilling, we may need a global file handle cache
that controls the number of open files. The same issue arises in the sort merge
phase: all spill files maintain an open file handle; we might exceed some limit.
For this PR, consider how long the file handle is open. Perhaps we need to
file a JIRA about managing the total number of open files.
> Support Spill to Disk for the Hash Aggregate Operator
> -----------------------------------------------------
>
> Key: DRILL-5457
> URL: https://issues.apache.org/jira/browse/DRILL-5457
> Project: Apache Drill
> Issue Type: Improvement
> Components: Execution - Relational Operators
> Affects Versions: 1.10.0
> Reporter: Boaz Ben-Zvi
> Assignee: Boaz Ben-Zvi
> Fix For: 1.11.0
>
>
> Support gradual spilling memory to disk as the available memory gets too
> small to allow in memory work for the Hash Aggregate Operator.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
