Hi,
I am thinking about the subject of installing one JEE application
multiple times in one server. An example would be to deploy 3 times:
myapp-dev, myapp-test and myapp-demo. First one would be used by CI on
every commit, to build, test, deploy, e2e. Second instance would be
for our testers, they will decide when to deploy. Third one for our
customer, so they can test it and give us feedback.
As of now I am developing on TomEE and I failed to achieve this goal.
I have ended up setting 3 distinct TomEE instances, it is pain and
very inconvenient.
This is all trivial when using Spring, because Spring apps lives their
own lives on application server. They do have their own isolated
contexts, they manage database connection pools, etc. When developing
JEE application, this is not the case any more. For example, there is
a datasource configuration in application server. One configures
jdbc/myapp in app server and then reference it everywhere using
@Resource(nape="jdbc/myapp"). This is cool sometimes, for example,
when application is spread across several, separated modules/webapps,
they all share single datasource. But what if I would like to start
another instance, like app-demo? As far as I know, each AS has its own
solutions for this (or does not have working solution). Glassfish has
glassfish-web.xml file, which can include this:
<resource-ref>
<res-ref-name>jdbc/myapp</res-ref-name>
<jndi-name>jdbc/myapp-demo</jndi-name>
</resource-ref>
Myapp-dev will have this file with:
<resource-ref>
<res-ref-name>jdbc/myapp</res-ref-name>
<jndi-name>jdbc/myapp-dev</jndi-name>
</resource-ref>
So the same code with @Resource annotation will refer to
jdbc/myapp-dev for development build and jdbc/myapp-demo for demo
build.
I did never test it though. Am I correct with the above? What about
other JNDI stuff, like bean names? Are they going to register
themselves in distinct locations for each application instance? How is
it going to work when BeanOne injects BeanTwo, both of them registered
three times by three instances of the same application?
Do I need to prepare my app for this scenario or everything (expect
for the resource-ref described above) will work out-of-the-box?
Thanks in advance, for any hints, tips and tricks,
Witold Szczerba
On 17 April 2013 00:04, Kevin Schmidt <kevin_at_nextgate.com> wrote:
> You can deploy the same WAR file multiple times with different names/context
> roots to accomplish what you want. You would need to do something like
> having the application look at its name/context root to load a different set
> of properties or use a different data source name to connect to the
> different database, but that is very straight forward.
>
> Note that this does in fact deploy the same code twice though. If you want
> only a single deploy but to use multiple backend databases, you may just
> need to make your app multi-tenant and have the logic in your code to
> connect to different databases based on the user.
>
> From: Blake McBride <blake_at_arahant.com>
> Reply-To: "users_at_glassfish.java.net" <users_at_glassfish.java.net>
> Date: Tue, 16 Apr 2013 14:38:13 -0700
> To: "users_at_glassfish.java.net" <users_at_glassfish.java.net>
> Subject: Can I run multiple instances of one application
>
> Greetings,
>
> I have a web application that connects to a single database and serves X
> number of simultaneous users without a problem.
>
> I would like to start up another instance of the same application that would
> connect to a different database and serve Y simultaneous users from a
> different URL.
>
> If this is possible with glassfish, presumably it would be able to use a
> single copy of the code base with independent and isolated data spaces.
>
> All the applications instances (for now) would be run on the same machine.
>
> I know I can just copy the application code, reconfigure for another
> database, and boot it as a separate application. The point is not have to
> have multiple copies of the application on the disk or in memory. The only
> things different between the instances would be the database it points to,
> the URL to access it, and the memory it takes to support the runtime data.
>
> I need this because the application is very large (15,000+ classes). I want
> to achieve the ability of creating many independent instances running
> without the need to have multiple copies of the code both on disk and in
> memory.
>
> Thanks.
>
> Blake McBride
>
>