Github user mateiz commented on a diff in the pull request:
https://github.com/apache/spark/pull/3003#discussion_r19620795
--- Diff: core/src/main/scala/org/apache/spark/executor/Executor.scala ---
@@ -210,25 +213,26 @@ private[spark] class Executor(
val resultSize = serializedDirectResult.limit
// directSend = sending directly back to the driver
- val (serializedResult, directSend) = {
- if (resultSize >= akkaFrameSize - AkkaUtils.reservedSizeBytes) {
+ val serializedResult = {
+ if (resultSize > maxResultSize) {
+ logInfo(s"Finished $taskName (TID $taskId). result is too
large (${resultSize} bytes),"
+ + " drop it")
+ ser.serialize(new TooLargeTaskResult(resultSize))
--- End diff --
It's kind of weird that we send a new type of result (TooLargeTaskResult)
in this case instead of failing the task. On the web UI, the task will look
like it completed, but then the job will abort. Is there a way we could fail
the task instead, and send a special TaskFailedReason that says the result is
too large? Then the TaskSetManager can abort when it sees that, the same way it
does for FetchFailed.
---
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 [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
