that is expected. it is up to the db server to notice the dropped tcp connection and clean it up. there is some good information at: https://issues.apache.org/jira/browse/AMQ-3654
I have been working on a variation of the default database locker that uses a time based lease rather than an open transaction. The idea being that a network glitch or db maintenance failover within the lease time need not require a master to stop. It will be used in conjunction with a JDBCIOExceptionHandler that will pause/resume the transport connections while the db is unavailable. see the relevant 5.7-SNAPSHOT config at: https://issues.apache.org/jira/browse/AMQ-3654?focusedCommentId=13294679&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-13294679 It would be great if that solves the lock expiry problem for you. or at least makes it more deterministic. On 13 June 2012 19:33, SergueiM <[email protected]> wrote: > With pure jdbc Master/Slave, when Master fails to connect to database (caused > by networking issue), Slave cannot acquire the lock. > Is this what one should expect? > > AMQ 5.6, Postrgres 9.1, Ubuntu. Each broker and DB are on separate hosts. > > Thank you. > - > Serguei > > -- > View this message in context: > http://activemq.2283324.n4.nabble.com/jdbc-Master-looses-DB-connection-Slave-won-t-acquire-the-lock-tp4653261.html > Sent from the ActiveMQ - User mailing list archive at Nabble.com. -- http://fusesource.com http://blog.garytully.com
