[BEAM-6103] Allow setting a HTTP write timeout to requests to BigQuery

• Type: Bug
• Status: Triage Needed
• Priority: P3
• Resolution: Fixed
• Affects Version/s: Not applicable
• Fix Version/s: Missing
• Component/s: io-java-gcp
• Assignee: Unassigned
• Reporter: Chamikara Madhusanka Jayalath
• Created: 20/Nov/18 19:45
• Updated: 13/Apr/23 11:03
• Resolved: 15/May/21 18:01

Description

Several users have reported that very occasionally some of the streaming insert requests to BigQuery are getting stuck. 

 

This can be mitigated by setting a write timeout for sockets and allowing runners to retry. But Java does not support setting write timeouts for non-NIO sockets [1] and this bug affects Google HTTP client library used by GCP IO.

 

Google HTTP library was recently updated to support write timeouts [2]. So we can update BigQueryIO to utilize this feature and allow setting a write timeout. Setting write timeout results in more threads being created. So this path can be slightly expensive and hence should not be made the default path.

[1] https://bugs.java.com/bugdatabase/view_bug.do?bug_id=4031100

[2] https://github.com/googleapis/google-http-java-client/releases/tag/v1.27.0

Issue Links

• links to: GitHub Pull Request #7097
• links to: GitHub Pull Request #12893