Github user mxm commented on a diff in the pull request:
https://github.com/apache/flink/pull/934#discussion_r35991960
--- Diff:
flink-runtime/src/main/java/org/apache/flink/runtime/executiongraph/ExecutionGraph.java
---
@@ -560,53 +621,177 @@ public ExecutionContext getExecutionContext() {
}
/**
+ * This works as cache for already merged accumulators, as, in some
cases,
+ * we do not want to remerge accumulators as this may lead to duplicate
entries.
+ */
+ private Map<String, Accumulator<?, ?>> mergedSmallUserAccumulators;
--- End diff --
Why can it lead to duplicate entries? The accumulators should always be
merged from scratch. Every accumulator update message contains the current
state of the accumulators.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
