Thanks a lot Fabrizio. With your details, now it is alined with what my thoughts were and what
jackrabbit few docs are saying.
It would be interesting to see how the other JCR170 implementations are behaving in these scenarios.
I know that at least the guys from jeceira are watching this list, so I will wait for their opinions
too.
I will try to find out more about these from exo and crx implementations. Do you guys know any other
implementations?
./alex
--
.the_mindstorm.
#: Fabrizio Giustina changed the world a bit at a time by saying on 9/26/2005
12:49 AM :#
Probably I was not clear: I simply wanted to highlight that you can't
have MULTIPLE instances of jackrabbit accessing to the same store
(filesystem, db, etc). This will corrupt your data and will cause
immediate problems due to caching (I was talking about jackrabbit
internal caching, that can't be flushed/disabled). This is what
happens if you start multiple instances of magnolia with the same
repository configuration (model 1)
You CAN share a repository, but you need a SINGLE instance of a jcr
repository active, and access it from different magnolia instances.
This can easily be done by accessing a single repository using jndi,
if running in the same application server instance (model 2).
If you plan to to use the same repository from different servers you
will need to access the jcr instance using RMI (there could be other
solution, this is the most immediate probably)... that means
implementing model 3.
I think this discussion could find more detailed answers on the
jackrabbit mailing list, this is actually not a Magnolia thing, but it
depends from the jcr implementation...
> 1) Setup author instance(s) to publish to a single hidden instance of
> Magnolia. The hidden instance uses a repository on a NFS exported
> file system.
>
> 2) Several public instances on different machines are configured to
> use a single repository on an NFS-mounted drive. Nothing will ever
> publish to these instances directly.
I tried something similar some time ago due to a wrong configuration,
publishing to a first instance and trying to access the same
repository store from another "read only" instance.
No, it will not work: errors everywhere due to data partially
persisted on the filesystem and partially cached... it may works for a
while, but you probably will end up with a corrupted repository... try
it at your risk
fabrizio
On 9/26/05, Alexandru Popescu <user-list@magnolia.info> wrote:
Hi Thomas!
This is indeed a very interesting architectural discussion. See my points
inlined:
#: Thomas M. Duffey changed the world a bit at a time by saying on 9/25/2005
11:31 PM :#
----------------------------------------------------------------
for list details see
http://www.magnolia.info/en/magnolia/developer.html
----------------------------------------------------------------
----------------------------------------------------------------
for list details see
http://www.magnolia.info/en/magnolia/developer.html
----------------------------------------------------------------
