[
https://issues.apache.org/jira/browse/BEAM-3061?focusedWorklogId=260351&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-260351
]
ASF GitHub Bot logged work on BEAM-3061:
----------------------------------------
Author: ASF GitHub Bot
Created on: 14/Jun/19 12:06
Start Date: 14/Jun/19 12:06
Worklog Time Spent: 10m
Work Description: steveniemitz commented on issue #7805: [BEAM-3061] Done
notifications for BigtableIO.Write
URL: https://github.com/apache/beam/pull/7805#issuecomment-502083671
> BTW will it make sense to group these objects so that we only output one
per window not one per batch ? Also it might also make sense to introduce
something like 'SpannerWriteResult' so that we only output these objects if
users explicitly ask for additional output. (specially if we do grouping due to
the cost of additional shuffle).
I don't think it's worth adding the GBK directly into the transform. Users
can either:
- Use Wait.on directly with this, which already does a shuffle itself
- Use the output directly, to handle when each bundle is committed
- Can just add a Sample.any themselves to get only a single firing when the
window closes.
Adding the GBK in the transform just removes flexibility for no real gain in
simplicity imo.
Also, there have been a couple discussions on the mailing list about this:
https://lists.apache.org/thread.html/ddcdf93604396b1cbcacdff49aba60817dc90ee7c8434725ea0d26c0@%3Cuser.beam.apache.org%3E
https://lists.apache.org/thread.html/8d5970c101c14c5b85fe0d3b53aaab169ab067ceb84941f4f69e4f44@%3Cdev.beam.apache.org%3E
I'm not sure if there is a "one size fits all" solution, because many IOs
operate differently (some can fail single elements, some rewindow into the
global window, etc), but I agree it'd be good to try to set a standard.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to 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: 260351)
Time Spent: 7h 50m (was: 7h 40m)
> BigtableIO should support emitting a sentinel "done" value when a bundle
> completes
> ----------------------------------------------------------------------------------
>
> Key: BEAM-3061
> URL: https://issues.apache.org/jira/browse/BEAM-3061
> Project: Beam
> Issue Type: Improvement
> Components: io-java-gcp
> Reporter: Steve Niemitz
> Assignee: Steve Niemitz
> Priority: Major
> Time Spent: 7h 50m
> Remaining Estimate: 0h
>
> There was some discussion of this on the dev@ mailing list [1]. This
> approach was taken based on discussion there.
> [1]
> https://lists.apache.org/thread.html/949b33782f722a9000c9bf9e37042739c6fd0927589b99752b78d7bd@%3Cdev.beam.apache.org%3E
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
