On 2/27/07, spiderman2 <[EMAIL PROTECTED]> wrote:

I've read the
http://activemq.apache.org/shared-file-system-master-slave.html doc  but the
diagram mentions a database. I thought we're only talking about writing to a
file system?? Just a typo?

We just reused the same images for both database and file based  master/slave.

BTW the website is a wiki so its very easy to fix typos...
http://activemq.apache.org/how-does-the-website-work.html


I see that the configuration mentioned in the doc is pointing to a file
system:
<journaledJDBC dataDirectory="/sharedFileSystem/broker"/>

but is called a "journaledJDBC." Does it also point to a default datasource
(db) or only a disk?

By default is uses Apache Derby using files in the same persistence
location. In 5.x this will use the AMQStore by default instead.


Question 2:
If 2 brokers are writing to one shared repository (be it db or file), then
there is still one point of failure. If the disk dies, I loose my high
availability.

Yes. The idea is you use a shared file system (like a SAN) to avoid
this. Or use a clustered database.


  -The only scenario I can find is Pure Master Slave that does replication.
But it has no failback mechanism.

It supports failover from the master to slave.

What are my activemq options for actual high availability with no
one-point-of-failure?

This page lists the 3 options with different trade-offs

http://activemq.apache.org/masterslave.html

--

James
-------
http://radio.weblogs.com/0112098/

Reply via email to