[
https://issues.apache.org/jira/browse/BEAM-4565?focusedWorklogId=117903&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-117903
]
ASF GitHub Bot logged work on BEAM-4565:
----------------------------------------
Author: ASF GitHub Bot
Created on: 30/Jun/18 20:34
Start Date: 30/Jun/18 20:34
Worklog Time Spent: 10m
Work Description: katsiapis commented on a change in pull request #5649:
[BEAM-4565] Fix hot key fanout in the face of combiner lifting.
URL: https://github.com/apache/beam/pull/5649#discussion_r199331674
##########
File path: sdks/python/apache_beam/transforms/core.py
##########
@@ -1118,7 +1118,7 @@ def _clone(self, **extra_attributes):
return clone
def with_fanout(self, fanout):
- return self._clone(fanout=fanout)
+ return self._clone(fanout=self.fanout)
Review comment:
We don't want "self." in front of the fanout here (since it is now being
ignored).
The test still pass of course, but the "optimization" (ie the fanout itself)
is lost.
My other comment was referring to a separate "self.fanout" issue, which has
however been fixed in https://github.com/apache/beam/pull//5800.
Could we fix the issue here?
/cc @aaltay
----------------------------------------------------------------
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]
Issue Time Tracking
-------------------
Worklog Id: (was: 117903)
Time Spent: 1h (was: 50m)
> Hot key fanout should not distribute keys to all shards.
> --------------------------------------------------------
>
> Key: BEAM-4565
> URL: https://issues.apache.org/jira/browse/BEAM-4565
> Project: Beam
> Issue Type: Task
> Components: sdk-java-core, sdk-py-core
> Affects Versions: 2.0.0, 2.1.0, 2.2.0, 2.3.0, 2.4.0, 2.5.0
> Reporter: Robert Bradshaw
> Priority: Major
> Time Spent: 1h
> Remaining Estimate: 0h
>
> The goal is to reduce the number of value sent to a single post-GBK worker.
> If combiner lifting happens, each bundle will sends a single value per
> sub-key, causing an N-fold blowup in shuffle data and N reducers with the
> same amount of data to consume as the single reducer in the non-fanout case.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
