> It kind of seems like there are two options. Option
> one is two individual servers running behind a load
> balancing router that you can upgrade the software
> one server at a time. However, this comes without
> session fail over I believe. Option 2 is a clustered
> appserver pair running behind a load balancing router
> that you have to upgrade software simultaneously
> thereby creating down time, but you get session fail
> over with this option. Is this correct?
Out of the box, yes I think this is correct.
Note, you can sort of mitigate the first problem by "using your own sessions". That is, assuming you don't use container based security (which relies on sessions), then "switching" to your own "session" implementation that relies on an off the shelf "clustered caching" solution as your "session" can mitigate the independent instances running without failover. Yes, you're effectively punting around the whole Servlet Session API and the container to do this, BUT, especially if it's all your own code, it may not be as traumatic of a change as you might think. Really depends on how important the "don't bounce folks during upgrades" requirement is.
> Other question, can GF be configured to use Oracle
> for HADB or do you have to use the HADB database? We
> have an Oracle RAC database behind the entire thing
> and would prefer to use it if possible.
I simply don't know having no experience with HADB. Sorry.
[Message sent by forum member 'whartung' (whartung)]
http://forums.java.net/jive/thread.jspa?messageID=308331