[
https://issues.redhat.com/browse/JGRP-2482?page=com.atlassian.jira.plugin...
]
Bela Ban commented on JGRP-2482:
--------------------------------
When running {{UPerf}} with {{<CLEAR_FLAGS internal="true"
oob="true"/>}} (which makes *all* messages regular), then the largest thread
pool size is ~5.
This points to {{OOB}} and/or {{INTERNAL}} messages being the culprit: these are forwarded
to the thread pool by {{MaxOneThreadPerSender}}, regardless of sender.
Next step: find out what OOB or INTERNAL messages are received (UNICAST3 ACKs?) and see if
the high number of these messages is correct.
Check if the largest thread pool is cluster-size + delta
--------------------------------------------------------
Key: JGRP-2482
URL:
https://issues.redhat.com/browse/JGRP-2482
Project: JGroups
Issue Type: Task
Reporter: Bela Ban
Assignee: Bela Ban
Priority: Major
Fix For: 5.0.0.Final
When running UPerf in a cluster of 4, we see {{thread_pool_size_largest=39}}. Verify that
MaxOneThreadPerSender works as expected: the largest thread pool size should be 4+delta
(timer threads, retransmission): ~6.
It looks as if {{MPerf}} behaves (~6 threads), but {{UPerf}} doesn't: ~26 threads
--
This message was sent by Atlassian Jira
(v7.13.8#713008)