[ 
https://issues.apache.org/jira/browse/BEAM-7192?focusedWorklogId=236203&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-236203
 ]

ASF GitHub Bot logged work on BEAM-7192:
----------------------------------------

                Author: ASF GitHub Bot
            Created on: 02/May/19 10:11
            Start Date: 02/May/19 10:11
    Worklog Time Spent: 10m 
      Work Description: mxm commented on issue #8441: [BEAM-7192] Fix 
partitioning of buffered elements during checkpointing
URL: https://github.com/apache/beam/pull/8441#issuecomment-488620160
 
 
   There are two threads:
   
   1. Main operator thread which emits elements and buffers them in the state 
backend
   2. GRPC thread which delegates state requests of pending elements to the 
state backend
   
   So there can be concurrent writes to the state backend which must be 
avoided. We don't need the lock in case only the main operator thread can have 
access, as it is the case for `flushBuffer()` where the bundle is ensured to be 
finished. That's why there is only a lock in the `buffer` method.
    
 
----------------------------------------------------------------
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: 236203)
    Time Spent: 1h 20m  (was: 1h 10m)

> Elements buffered during checkpointing may not be partitioned correctly
> -----------------------------------------------------------------------
>
>                 Key: BEAM-7192
>                 URL: https://issues.apache.org/jira/browse/BEAM-7192
>             Project: Beam
>          Issue Type: Bug
>          Components: runner-flink
>            Reporter: Maximilian Michels
>            Assignee: Maximilian Michels
>            Priority: Major
>             Fix For: 2.13.0
>
>          Time Spent: 1h 20m
>  Remaining Estimate: 0h
>
> When a Flink checkpoint is taken, the current bundle is finalized. The 
> finalization happens when the checkpoint barrier has already been sent 
> downstream; emitting elements at this point would violate the checkpoint 
> barrier alignment.
> When elements are emitted during checkpointing they are buffered until the 
> checkpoint is complete. We should ensure that they are keyed correctly and 
> emission of the buffered elements does not interfere with any concurrent 
> state requests (in case of portability).
>  



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to