[FLINK-31396] Occasional inaccurate timeout time calculation with System.nanotime in batch read buffer pool

• Type: Bug
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 1.18.0
• Fix Version/s: 1.18.0
• Component/s: Runtime / Network
• Labels: pull-request-available
• Assignee: Yuxin Tan
• Reporter: Yuxin Tan
• Created: 10/Mar/23 08:29
• Updated: 10/Mar/23 14:40
• Resolved: 10/Mar/23 14:40

Description

When running TPC-DS tests, I encountered the read buffer request timeout because of configuring too less read buffers. But I found the timeout time may be less than 5m occasionally, 5m is the expected time.
I read the docs of System.nanotime, the docs say that t1 < t0 should not be used, because of the possibility of numerical overflow. I tested the System.currentTimeMillis and it can work as expected.

Issue Links

• links to: GitHub Pull Request #22151