You're right, I read that JDBC Persistance doesn't work with fast journalling. But I can't understand why not? It sounds like it would be an ideal HA scenario.
The 3 Master Slave Options: 1) Pure: Doesn't support an automatic failback - meaning after a failure I'll only have 1 Broker who may eventually fail and my system will die. 2) JDBC - Too slow. I'm getting rates of 200ms/ synchronous message. (Altho i will explore transactions under async message) 3) SAN - which is an option but expensive!! I really like the Fast Journaling, but is it not supported in any HA scenario?? Does anyone have a recommended deployment of HA with minimal to no message loss (that isn't as slow as JDBC persistance)? James.Strachan wrote: > > On 2/22/07, spiderman2 <[EMAIL PROTECTED]> wrote: >> >> I'm looking into clustered HA deployments with zero message loss under >> one >> point-of-failure. >> >> My best options seem to be: >> A) Network of Brokers with fast journaling and jdbc persistance > > Networks of brokers are for store and forward, not for HA of messages. > i.e. a message only exists on one broker in a network at any point in > time - its never HA across brokers > > >> B) JDBC Master Slave > > Yes - the 3 master/slave alternatives are what you need > http://activemq.apache.org/masterslave.html > > >> BUT, what if broker dies before it writes its batch to the database, but >> after it completes its fast-journal? > > It depends which of the 3 master/slave alternatives you are using - > but with JDBC Master/Slave you don't use the journal > > > -- > > James > ------- > http://radio.weblogs.com/0112098/ > > -- View this message in context: http://www.nabble.com/Recover-a-failed-Broker%27s-Journal-%28by-Database-of-Another-Broker%29-tf3275595s2354.html#a9119970 Sent from the ActiveMQ - User mailing list archive at Nabble.com.
