[ https://issues.jboss.org/browse/JGRP-1663?page=com.atlassian.jira.plugin.... ] Bela Ban commented on JGRP-1663: -------------------------------- That's definitely not the way TCPPING shuld be used ! I suggest use a different discovery protocol, e.g. TCPGOSSIP or - if multicasting is enabled - MPING. I also suggest to set TCP.sock_conn_timeout to 200 ms (default:2s), so that TCP connection attempts to down hosts are not filling up the timer thread pool. TCPPING does its discovery in parallel, so each of the discovery requests causes 50+ timer tasks to be added to the timer. There's something else going on though as this should not accumulate 19'000 threads. The stack trace you pasted above shows threads waiting for the lock on connection establishment, do you see which thread actually acquired the lock and whether it is blocked itself ? Also: STABLE.desired_avg_gossip="10": that's causing a STABLE multicast from *every* member every 10ms ! That's not the default and this causes *a lot of traffic*, possibly causing connection establishment to block. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira