Erik Salter created ISPN-2423: --------------------------------- Summary: Handle the site unreachable case when sending unicasts to backup sites. Key: ISPN-2423 URL: https://issues.jboss.org/browse/ISPN-2423 Project: Infinispan Issue Type: Bug Components: Cross-Site Replication Affects Versions: 5.2.0.Beta2 Reporter: Erik Salter Assignee: Mircea Markus When the BackupInterceptors send a unicast to an unreachable site, the underlying RELAY2 protocol can be configured to retry sending the message. The current implementation iterates over sites and sequentially tries to send a message. If one of the sites is unreachable, it will block waiting for a response. For instance, if in sites {A,B,C} and A is down, we can block on waiting for the response from A. As this is in the critical path for transactions, locks, etc., this affects the overall throughput of the system. Note that this happens for async and sync replication. A potential solution is in JGroupsTransport.backupRemotely(), we create a Runnable which calls MessageDispatcher.sendWithFuture() and submit it to the internal asyncExecutor thread pool. For async replication, the responses can then be processed (and sites taken offline, etc) outside the tx critical path. -- 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