GitHub user cloud-fan opened a pull request:
https://github.com/apache/spark/pull/22907
[SPARK-25896][CORE][WIP] Accumulator should only be updated once for each
successful task in shuffle map stage
## What changes were proposed in this pull request?
This is a followup of https://github.com/apache/spark/pull/19877
For the same reason, we should also update accumulator once for each
successful task in shuffle map stage.
TODO:
1. `ShuffleMapStage` has `pendingPartitions` and `findMissingPartitions`,
I'm not sure which one is the single source of truth
2. When we receive repeated successful shuffle map tasks, seems we will
override the previous one. Need a double check.
3. add tests.
## How was this patch tested?
TODO
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/cloud-fan/spark accum
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/spark/pull/22907.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #22907
----
commit 7876d9c44c15fe1fb64f1d7587c97e23ff2be5a2
Author: Wenchen Fan <wenchen@...>
Date: 2018-10-31T15:23:48Z
Accumulator should only be updated once for each successful task in shuffle
map stage
----
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
