Hi, we have the following problem with ActiveMQ: we have some consumer code running in an app server and acknowledging messages manually. ActiveMQ broker is running outside the app server as a separate process. Everything is ok until we restart the broker. After we restart it resends several (a few) messages acknowledged hours ago (goes back to 12-14 hours). There are not many: yesterday we had 3 from many thousands at that day. These 3 messages were consumed at different times of the day. KahaDB is running with enableJournalDiskSyncs=false. We also observed that each time such a message was consumed and acknowledged KahaDB rotated the data file. So my suspicios is that there is somewhere a leak in KahaDB logic when using async writes and rotating the data files. So a KahaDB remove command is not persisted and after ActiveMQ restart, the broker rereads commands from the files and resends the corresponding messages. Is it a feasible explanation? Does it make sense? Maybe there is already some JIRA from this area? I did not find any clues during my initial analysis of ActiveMQ source code. Any help appreciated. Currenty we are planing to switch on journal syncs with the hope that this will work around this problem.
Best regards, Sergiy -- View this message in context: http://activemq.2283324.n4.nabble.com/ActiveMQ-5-10-looses-message-acknowledgements-tp4690306.html Sent from the ActiveMQ - User mailing list archive at Nabble.com.
