[FLINK-28240] NettyShuffleMetricFactory#RequestedMemoryUsageMetric#getValue may throw ArithmeticException when the total segments of NetworkBufferPool is 0

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.15.0
• Fix Version/s: 1.16.0, 1.15.2
• Component/s: Runtime / Network
• Labels: pull-request-available
• Assignee: Gen Luo
• Reporter: Gen Luo
• Created: 24/Jun/22 07:59
• Updated: 05/Jul/22 07:12
• Resolved: 05/Jul/22 07:12

Description

In a single vertex job, the network memory can be set to 0 since the job doesn't need it, and in this case the totalNumberOfMemorySegments of the NetworkBufferPool will also be 0.

While the NettyShuffleMetricFactory#RequestedMemoryUsageMetric#getValue uses the totalNumberOfMemorySegments of NetworkBufferPool as divisor without validating, so an ArithmeticException will be thrown when the totalNumberOfMemorySegments is 0.

Since 0 network memory is in fact valid for a single vertex job, I suppose the RequestedMemoryUsageMetric#getValue should check if the devisor is 0, and return 0 as the usage directly in such cases.

Issue Links

• links to: GitHub Pull Request #20108