[
https://issues.apache.org/jira/browse/BEAM-14437?focusedWorklogId=769239&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-769239
]
ASF GitHub Bot logged work on BEAM-14437:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 11/May/22 18:52
Start Date: 11/May/22 18:52
Worklog Time Spent: 10m
Work Description: kennknowles commented on PR #17585:
URL: https://github.com/apache/beam/pull/17585#issuecomment-1124176469
Is this going to be a safe default across lots of different contexts? This
same code runs locally, during testing, on Spark & Flink & Samza, etc.
Issue Time Tracking
-------------------
Worklog Id: (was: 769239)
Time Spent: 1h (was: 50m)
> Raise Java Http KeepAlive Cache Size
> ------------------------------------
>
> Key: BEAM-14437
> URL: https://issues.apache.org/jira/browse/BEAM-14437
> Project: Beam
> Issue Type: Improvement
> Components: runner-dataflow
> Reporter: Arun Pandian
> Assignee: Arun Pandian
> Priority: P2
> Time Spent: 1h
> Remaining Estimate: 0h
>
> Java HTTP[S]Connection's keep alive by default keeps 5 idle connections
> alive. Dataflow user workers could make 1000s of connection to the same
> endpoint. The HTTPS handshake overhead is significant in pipelines using
> BigQueryIO streaming inserts and bumping up the keep alive limit reduces it.
> The keep alive connection count can be changed by setting the JVM system
> property "http.maxConnections".
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
