On 5/10/06, Dave cawthorn <[EMAIL PROTECTED]> wrote:
Thanks James,
I did a quick test of setting a Sequoia connection's autocommit property to
false using
con.setAutoCommit(false); and both commits and rollbacks seem to work.
Hmm - autocommit seems to work fine for other databases though.
I will go through the activemq source to see if i can see any other reason
why its bombing.
I was going to have a look at master/slave but from reading the forum there
is no way to resume a master without stopping the slave and copying its data
back to the master. So correct me if I'm wrong but eventually you have to
down the service cause you only get one failover?
Yes - though the clients can auto-reconnect to another broker. So you
could bring a new broker up, let folks use that - then bring the old
master back on line when its all in sync again - then you have
continuous availability and no message loss.
The only missing feature is we don't have automatic old-master <->
slave-now-a-master synchronisation. Thats a manual process right now.
I'm currently planning on using a replicated message store and having
activemq as a hasingleton in a jboss cluster. This way only one broker will
ever be active but messages will be replicated on two machines and I will be
able to "fail back" to the first node if the second node fails because i
have the ability to resync the masters message store when it comes back up.
Cool
PS message groups do rock!
I *love* messsage groups!
--
James
-------
http://radio.weblogs.com/0112098/
