Skip to content

ARTEMIS-6355 perf client back on nanoTime, and a negative latency counted and reported instead of recorded as 0 - #6909

Open
grepjava wants to merge 1 commit into
apache:mainfrom
grepjava:ARTEMIS-6355-nanotime
Open

grepjava wants to merge 1 commit into
apache:mainfrom
grepjava:ARTEMIS-6355-nanotime

Conversation

@grepjava

Copy link
Copy Markdown
Contributor

Follow-up to #6873, after the review there.

  • perf client goes back to System.nanoTime(): it only receives the messages it sent itself, in the same process, and nanoTime does not step back or lose precision without --add-opens. perf producer and perf consumer keep the wall clock they share.
  • A negative latency is no longer recorded as 0. The receiving listener and the send-completion path count it instead, and the summary warns how many there were and that the latency results are not valid; the JSON summary carries the count as negativeLatencies.

MessageListenerBenchmarkTest checks that a message stamped ahead of the clock is counted rather than recorded, and that the summary warns only when there are negative latencies.

https://issues.apache.org/jira/browse/ARTEMIS-6355

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant