[ https://issues.jboss.org/browse/JGRP-1809?page=com.atlassian.jira.plugin.... ] Bela Ban updated JGRP-1809: --------------------------- Description: Sometimes tests in the testsuite fail because the nodes run over SHARED_LOOPBACK fail to discover each other and therefore don't form a cluster. This can happen because discovery requests sent via PING are discarded on a full thread pool. Because the default test stack doesn't have a MERGE protocol, members won't get merged back. h5. GOAL * Create a discovery protocol which always discovers other members registered with the same cluster h5. SOLUTION * Instead of using SHARED_LOOPBACK to send and receive discovery requests, query SHARED_LOOPBACK directly for discovery information. E.g. if we have members A,B,C registered with cluster "demo" in SHARED_LOOPBACK, then a node D joining can fetch (via SHARED_LOOPBACK_PING) the PingData of A, B and C directly from SHARED_LOOPBACK. * This way, we won't have any failed test cases (run over SHARED_LOOPBACK) which fail because no other members were discovered. Also, discovery (even for the first node started in a cluster) is very fast, and we don't need the timeout at all. * Also, we don't need to add a MERGE protocol to the stack as merging is not needed was: Sometimes tests in the testsuite fail because the nodes run over SHARED_LOOPBACK fail to discover each other and therefore don't form a cluster. This can happen because discovery requests sent via PING are discarded on a full thread pool. Because the default test stack doesn't have a MERGE protocol, members won't get merged back. h5. GOAL * Create a discovery protocol which always discovers other members registered with the same cluster h5. SOLUTION * Instead of using SHARED_LOOPBACK to send and receive discovery requests, query SHARED_LOOPBACK directly for discovery information. E.g. if we have members A,B,C registered with cluster "demo" in SHARED_LOOPBACK, then a node D joining can fetch (via SHARED_LOOPBACK_PING) the PingData of A, B and C directly from SHARED_LOOPBACK. This way, we won't have any failed test cases (run over SHARED_LOOPBACK) which fail because no other members were discovered. Also, discovery (even for the first node started in a cluster) is very fast, and we don't need the timeout at all. -- 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