If I have a pure master-slave configuration and the master fails, the recovery is to:
- shut down the master (if it's not already dead) - shut down the slave - copy "the contents of the data directory" from the slave to the master - start up the master Question 1: In the data directory, there are 2 main subdirectories (along with the log files, which wouldn't get copied): the kahadb subdirectory, and a subdirectory named for the broker itself (in my case, mq01.a, or mq02.a for the slave) Does the broker-named directory need to be copied too? If so, should it be renamed to the same as the master, so that the master will pick it up? Question 2: After starting the master, should I clean out the slave directory? Reconnecting the slave after the master has been active for a bit gives me errors in the master log file about stale subscriptions when the slave connects; but cleaning out the slave data dir would mean that any persistent messages saved there from before the master crashed would now only exist on the master, because they won't get synced across when the slave connects; only new messages get persisted. -- View this message in context: http://activemq.2283324.n4.nabble.com/Pure-master-slave-recovery-questions-tp4216601p4216601.html Sent from the ActiveMQ - User mailing list archive at Nabble.com.
