Well, keep in mind that with the discontinuation of pure master/slave, persistence is the only way to get true HA. So your point about being able to withstand an outage of part of your datacenter is possible only because you have persistence in the first place. (And yes, LevelDB is not giving you any of those guarantees for any non-persistent messages that your producers might send.) On Feb 2, 2015 4:19 PM, "Kevin Burton" <[email protected]> wrote:
> Given that: > > 1. ActiveMQ shouldn’t be your main database. You should have a secondary > system to store message and send them to AMQ > > 2. You may have repair code to re-push messages into AMQ > > 3. you’re running with LevelDB replication… > > .. then what’s the purpose of persistence? > > It really only handles the situation where you have an ENTIRE datacenter > failure. If you have one node fail, the messages should be replicated. > > And if you DO have a datacenter failure, you can just repair your queue > from your database. > > Of course, I don’t *think* that leveldb replicates non-persistent messages. > > -- > > Founder/CEO Spinn3r.com > Location: *San Francisco, CA* > blog: http://burtonator.wordpress.com > … or check out my Google+ profile > <https://plus.google.com/102718274791889610666/posts> > <http://spinn3r.com> >
