vatanrathi opened a new issue, #25991: URL: https://github.com/apache/beam/issues/25991
### What happened? Beam java pipeline encounter severe performance issues on all beam versions later than 2.30.0. After troubleshooting, we identified that enhancement done under BEAM-12329 seems root cause for the performance degradation. Note that pipeline runs fine for small datasets of few MBs but are significant slower for lager datasets ~10GBs+. It appears that on version 2.30 and lower, amazon-web-service module was not draining the input stream and program complained about "Not all bytes were read from .... " but as we know in this case system is overreacting and making invalid assumption about remaining data and cost of read vs abort and reconnect. In later version post 2.30.0, when input stream is drained when close function is called and subsequent requests to read remaining data needs to reconnect and seek to the position which is significantly slower in our case and pipeline keeps running for hours as against to a few minutes on version 2.30 and lower. Do you have any suggestion on how to achieve the same performance ? I have tried both amazon-web-services and amazon-web-services2 on all versions after 2.30 and all results same poor performance. ### Issue Priority Priority: 3 (minor) ### Issue Components - [ ] Component: Python SDK - [X] Component: Java SDK - [ ] Component: Go SDK - [ ] Component: Typescript SDK - [ ] Component: IO connector - [ ] Component: Beam examples - [ ] Component: Beam playground - [ ] Component: Beam katas - [ ] Component: Website - [X] Component: Spark Runner - [ ] Component: Flink Runner - [ ] Component: Samza Runner - [ ] Component: Twister2 Runner - [ ] Component: Hazelcast Jet Runner - [ ] Component: Google Cloud Dataflow Runner -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
