[
https://issues.apache.org/jira/browse/BEAM-8352?focusedWorklogId=326989&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-326989
]
ASF GitHub Bot logged work on BEAM-8352:
----------------------------------------
Author: ASF GitHub Bot
Created on: 11/Oct/19 16:53
Start Date: 11/Oct/19 16:53
Worklog Time Spent: 10m
Work Description: aromanenko-dev commented on issue #9745: [BEAM-8352]
Add "withMaxCapacityPerShard()" to KinesisIO.Read
URL: https://github.com/apache/beam/pull/9745#issuecomment-541141921
I agree that limitation based on payload size would be better, though it's
still unclear how we can manage it dynamically.
1. Should be initial value based on process heap size as default value or
configured by user?
2. How much of free space we should leave unused to prevent NPE because of
new arrived records (we don't know their size in advance)?
3. Do we need to adapt the maximum queue size in runtime?
4. If yes, what are the causes for that?
Btw, do you already have a solution for Java portability container problem?
Either maybe we have something similar implemented for other IO?
----------------------------------------------------------------
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: 326989)
Time Spent: 1h 40m (was: 1.5h)
> Reading records in background may lead to OOM errors
> ----------------------------------------------------
>
> Key: BEAM-8352
> URL: https://issues.apache.org/jira/browse/BEAM-8352
> Project: Beam
> Issue Type: Bug
> Components: io-java-kinesis
> Affects Versions: 2.2.0, 2.3.0, 2.4.0, 2.5.0, 2.6.0, 2.7.0, 2.8.0, 2.9.0,
> 2.10.0, 2.11.0, 2.12.0, 2.13.0, 2.14.0, 2.15.0
> Reporter: Mateusz Juraszek
> Assignee: Alexey Romanenko
> Priority: Major
> Fix For: 2.17.0
>
> Time Spent: 1h 40m
> Remaining Estimate: 0h
>
> We have faced a problem with OOM errors in our dataflow job containing
> Kinesis sources. After investigation, it occurred that the issue was caused
> by too many records being consumed by Kinesis sources that pipeline couldn't
> handle in time.
> Looking into the Kinesis connector's code, the internal queue (recordsQueue)
> that records are being put in the background is setup for each
> ShardReadersPool (created for each source being Kinesis stream). The size of
> the queue is set to `queueCapacityPerShard * number of shards`. The bigger
> number of shards, the bigger queue size. There is no ability to limit the
> maximum capacity of the queue (queueCapacityPerShard is also not configurable
> and it's set to DEFAULT_CAPACITY_PER_SHARD=10_000). Additionally, there is no
> differentiation on records size, so the size of data placed to the queue
> might increase to the point where OOM will be thrown.
> It would be great to have ability to somehow limit the number of records that
> are being read in the background to some sensible value. At the beginning,
> simple solution would be to allow configuring max queue size for source at
> the creation of KinesisIO.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
