>> All of ours are. We don't use clusters that span machines (yet), and we
>> don't (or hardly) use the DAS to run applications on.
>> So all production as well as tests run on a cluster with its one or
>> multiple instances on the same machine.
>
> Why are you running multiple instances on a single machine?
> Is it because the JVM doesn't scale well? Or does GlassFish
> itself not scale well enough to handle all the load in a single
> instance? Or perhaps you want to be able to survive the crash
> of a server process? (Does that happen often?)
Because:
- the JVM doesn't always scale well
- to be able to survive the server process (which I think doesn't happen
often, but we need to be prepared. One type of crash I sometimes hear
about is because of OOM)
- before we used GlassFish, we did it that way (multiple instances on
one machine) for years, for the above reasons, so we are comfortable and
experienced with such structure
> Are you running a load balancer on the same machine or on a separate
> machine?
We have both cases (but separate machine (in DMZ) is more common).
>> I'm planning to confirm what issues could arise on GFv2.x (because I
>> don't think there are any) with having the instances point to the DAS's
>> application folder directly instead of each having its own copy. If we
>> can do that in GFv3.1 by default, that should reduce sychronization time
>> during deployment and start-up for this case. And required disk space.
>
> I think applications might work. We should support applications that
> are "directory deployed" in a shared filesystem, and this would be similar.
I believe GFv2.x doesn't support the "directory deployed" function in
case of clusters. If we can make that work using a shared filesystem,
that would be great!
> Anything in the config directory would only work by luck, not by design.
Just the files under 'applications' is probably enough. They can be big
(or many small ones with web applications), while the files under config
don't increase and (except for domain.xml) probably never change anyway.
Thanks,
Dies
>> On 11/03/2010 03:38, Tim Quinn wrote:
>>> Does anyone have a sense for how many GlassFish environments are
>>> "cluster development" or "cluster testing" environments in which all of
>>> these activities - adding instances, deploying and redeploying apps -
>>> are more frequent than in truly production settings? And to the extent
>>> that we want to offer a compelling (that is, fast!) development
>>> experience with clusters as well, we'll want to keep those environments
>>> in mind.
>> <snip>