asdf2014 commented on a change in pull request #6767: Fix auto compaction to
consider intervals of running tasks
URL: https://github.com/apache/incubator-druid/pull/6767#discussion_r243871801
##########
File path:
server/src/main/java/org/apache/druid/server/coordinator/helper/NewestSegmentFirstIterator.java
##########
@@ -339,51 +345,134 @@ private static SegmentsToCompact findSegmentsToCompact(
/**
* Returns the initial searchInterval which is {@code
(timeline.first().start, timeline.last().end - skipOffset)}.
*
- * @param timeline timeline of a dataSource
- * @param skipOffset skipOFfset
+ * @param timeline timeline of a dataSource
+ * @param skipIntervals intervals to skip
*
* @return found interval to search or null if it's not found
*/
- @Nullable
- private static Interval findInitialSearchInterval(
+ private static List<Interval> findInitialSearchInterval(
VersionedIntervalTimeline<String, DataSegment> timeline,
- Period skipOffset
+ Period skipOffset,
+ @Nullable List<Interval> skipIntervals
)
{
Preconditions.checkArgument(timeline != null && !timeline.isEmpty(),
"timeline should not be null or empty");
Preconditions.checkNotNull(skipOffset, "skipOffset");
final TimelineObjectHolder<String, DataSegment> first =
Preconditions.checkNotNull(timeline.first(), "first");
final TimelineObjectHolder<String, DataSegment> last =
Preconditions.checkNotNull(timeline.last(), "last");
+ final List<Interval> fullSkipIntervals = sortAndAddSkipIntervalFromLatest(
+ last.getInterval().getEnd(),
+ skipOffset,
+ skipIntervals
+ );
- final Interval skipInterval = new Interval(skipOffset,
last.getInterval().getEnd());
+ final Interval totalInterval = new
Interval(first.getInterval().getStart(), last.getInterval().getEnd());
+ final List<Interval> filteredInterval = filterSkipIntervals(totalInterval,
fullSkipIntervals);
+ final List<Interval> searchIntervals = new ArrayList<>();
- final DateTime lookupStart = first.getInterval().getStart();
- final DateTime lookupEnd = last.getInterval().getEnd().minus(skipOffset);
- if (lookupStart.isBefore(lookupEnd)) {
+ for (Interval lookupInterval : filteredInterval) {
final List<TimelineObjectHolder<String, DataSegment>> holders =
timeline.lookup(
- new Interval(lookupStart, lookupEnd)
+ new Interval(lookupInterval.getStart(), lookupInterval.getEnd())
);
final List<DataSegment> segments = holders
.stream()
.flatMap(holder ->
StreamSupport.stream(holder.getObject().spliterator(), false))
.map(PartitionChunk::getObject)
- .filter(segment -> !segment.getInterval().overlaps(skipInterval))
+ .filter(segment -> lookupInterval.contains(segment.getInterval()))
.sorted((s1, s2) ->
Comparators.intervalsByStartThenEnd().compare(s1.getInterval(),
s2.getInterval()))
.collect(Collectors.toList());
- if (segments.isEmpty()) {
- return null;
- } else {
- return new Interval(
- segments.get(0).getInterval().getStart(),
- segments.get(segments.size() - 1).getInterval().getEnd()
+ if (!segments.isEmpty()) {
+ searchIntervals.add(
+ new Interval(
+ segments.get(0).getInterval().getStart(),
+ segments.get(segments.size() - 1).getInterval().getEnd()
+ )
);
}
+ }
+
+ return searchIntervals;
+ }
+
+ @VisibleForTesting
+ static List<Interval> sortAndAddSkipIntervalFromLatest(
+ DateTime latest,
+ Period skipOffset,
+ @Nullable List<Interval> skipIntervals
+ )
+ {
+ final List<Interval> nonNullSkipIntervals = skipIntervals == null
+ ? new ArrayList<>()
Review comment:
Maybe `new ArrayList<>(1)` would be better.
----------------------------------------------------------------
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]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]