Github user franz1981 commented on a diff in the pull request:
https://github.com/apache/activemq-artemis/pull/2206#discussion_r206849644
--- Diff:
artemis-server/src/main/java/org/apache/activemq/artemis/core/server/cluster/impl/BridgeImpl.java
---
@@ -609,6 +612,7 @@ public HandleStatus handle(final MessageReference ref)
throws Exception {
final Message message = beforeForward(ref.getMessage(), dest);
pendingAcks.countUp();
+ metrics.incrementEnqueueCount();
--- End diff --
I haven't had the chance to explain that in other PRs related to metrics
collection: if these metrics are getting updated from a single thread/writer,
instead of using `AtomicLong::incrementAndGet` can be used
`AtomicLong::lazySet(AtomicLong::get() + delta)` from the pov of metrics
collections it will be a lot less expensive.
It shouldn't be visible in our tests on laptop HW but with real servers,
especially with NUMA (and multi socket) configuration.
I don't know if it's the case, but given that it cost very little to change
that, I suggest to do it right where is possible to do it.
---
