On Jul 13, 2007, at 2:37 PM, Glen Mazza wrote:
Hello,
I am thinking of using the DerbyDB embedded in Geronimo for a web
application. I have a couple of questions about the Derby that
is in
Geronimo:
A couple of comments.
I think you are going to need an ear file or a geronimo "synthetic
ear" or at least 2 artifacts -- a rar for the database pool and your
web app. I think you may find an ear with "external modules" for the
db pool is the simplest: you can include the geronimo plan inside the
ear and have only one deployable. The daytrader app shows how to do
this in e.g the plans/daytrader-g-2.0.xml plan (I might not quite
have the name right) although that plan is not included in the ear.
Thanks. I will look into this. I want to actually go both routes:
1.) I am thinking of a read-only database with dummy data placed
into a WAR
file. I want to demonstrate XSL/XSLT report generation using any
sample
data. (Later, the user can modify the code to access any other
database
desired.) If I can place it in the WAR, then the webapp can
quickly access
it "out of the box" from *any* servlet container.
Last I heard (several years ago) derby didn't support read only
databases -- it needed somewhere to write locks or something like
that. Can derby work from a resource url?
2.) For a more meaningful read-write production usage, doing as
you are
saying above, and possibly documenting it on the Geronimo wiki so
others can
use it/add to it.
1.) Does Geronimo's Derby run in embedded mode or network server
mode? If
the former (which I believe but would like to confirm), does that
mean I
would need a separate instance of Derby if I want to run in Network
server
mode?
You can access derby from geronimo using either embedded or network,
depending on which tranql resource adapter you use. AFAIK embedded
is likely to be a bit faster.
I was not able to access it in client-server mode (accessing other
Derby
DB's external to Geronimo's probably would not have been a problem
though.)
Apparently the Derby database incorporated by Geronimo runs in
embedded
mode--only the apps from one JVM can access it--so it can't be
accessed
externally by Derby's "ij" query tool for example, unless Geronimo
itself is
shut down. (If you go to the "DB Info" page on the Geronimo
console, it
says the Derby Embedded Driver is being used.)
To test this, I tried to log into my sample Derby database I
created in
Geronimo (located in geronimo/var/derby/mydbname) from ij -- it
failed when
Geronimo was running, but worked after I shut down Geronimo. I did
this
last year with Derby standalone and documented it:
[1] http://www.jroller.com/page/gmazza?
entry=apache_derby_in_network_server
well, from this gbean in system-database:
<gbean name="DerbyNetwork"
class="org.apache.geronimo.derby.DerbyNetworkGBean">
<reference name="derbySystem"><name>DerbySystem</name></
reference>
<attribute name="host">localhost</attribute>
<attribute name="port">1527</attribute>
</gbean>
and the success of a bunch of tck tests that use app client access to
the derby instance running in the server, I'm pretty sure that the
network connector is turned on and working. Maybe the host=localhost
rather than 0.0.0.0 or a specific interface is the problem? What
connect string did you use for ij?
2.) Is it possible (or practical) to embed Derby within a web
application
WAR file, preconfigure it with sample tables and rows, etc., and
have the
webapp access *that* embedded Derby instance, and not the one that
comes
with Geronimo (or any network server version possibly running on the
machine)? I am asking this because it would be nice for a sample
web
application if I could configure it such that no database
configuration
would be necessary for a user; they would just need to deploy the
WAR and it
will work out-of-the-box.
I don't think its practical or a good idea to run the database
actually in the war file. For one thing it really ought to be read
only if you did that :-)
What I recommend is to construct your starting database, wrap it up
in a jar file, include the jar in the ear or war file, and include a
gbean in the geronimo plan to unpack the jar file into the var/derby
directory. Our intent is that all program modifiable data files such
as embedded databases should go somewhere in var.
That's a clean design.
We have an example of doing almost exactly this in geronimo, in the
uddi app.
applications/geronimo-uddi-db constructs the db and packs it in a jar
applications/geronimo-uddi-server constructs the war file
configs/uddi-jetty6 (and uddi-tomcat) "predeploy" uddi into the jetty
or tomcat server, constructing what is basically a geronimo plugin.
Thanks. I'll take a look at this.
Let us know if you have further problems!
thanks
david jencks
Glen
--
View this message in context: http://www.nabble.com/Derby-DB-in-
Geronimo-tf4068958s134.html#a11585231
Sent from the Apache Geronimo - Users mailing list archive at
Nabble.com.
