geronimo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Hernan Cunico <>
Subject Re: Some observations from a user perspective on deploying an application
Date Fri, 28 Apr 2006 01:23:05 GMT
Hi All,
today I got my first, and mostly functional, G 1.1 build and succeeded in deploying my first
data source through the console. I choose DB2 because of the multiple jars.

My impression at the end was, it is even more complicated than the previous version. The new
convention is not exactly "charming"! :)

I think it is most likely that copying the jars to the repo will be a "command line" task
since you 
already have to change (version) the name of the jars. Once you have the drivers in place
then you 
move to the console and try to deploy the data source.

Using the console to install the jars is certainly better than doing it one-by-one manually
still there are 5 fields that need to be filled/verified for each jar to be installed. I think
option of choice for installing (copying) the drivers would still be the command line unless
we can 
simplify the GUI process.

One way of simplifying it could be:

1- use the wizard to create a data source (first step as usual)
2- From the Driver Jar pull down menu...
      - it might be better to have a scroll down box where you can select (ctrl+) multiple
        jars if needed.
      - in the same scroll box have an option to install other drivers not listed
        other drivers not listed:
         - have a browse button to select multiple files to install in the same single operation
           if the files are already available locally (instead of going through the common
           libraries).  This option should keep to a minimum user
           required input.
           The multiple files selected serve one single purpose, hence it would be nice to
have them
           treated (listed) as a block, for example *DB2 8.1 JARs*
         - provide a download option (already available)
      - once the driver is installed continue with the regular existing procedure.
3- Optionally configure some of the connection pool parameters (timeouts, min and max pool
sz, etc)
4- Test and deploy

When editing an existing data source it would be nice to have the option for *Testing* the

connection again.


Matt Hogstrom wrote:
> All,
> I was looking at Aaron's "Guiding Principals" doc and rmembered I meant 
> to post some thoughts on the "Geronimo Experience."
> I've been working on deploying daytrader and have noted a couple of 
> things that I think are relevant from a user perspective.  Specifically 
> these around the repository.
> For my testing I've been using DB2 as the database as I wanted to use a 
> commercial database that was external to the AppServer.  The same story 
> is true for MySQL, PostGres, etc.
> In order to deploy a dependent jar into the repo you have to add the 
> group, artifact and version information.  So in the case of DB2 there 
> are two jars you have to add.  They are the db2jcc.jar and the 
> db2jcc_license_cu.jar.  Putting them into the repo I did the following:
> mkdir -p 
> ./geronimo-1.1-SNAPSHOT/repository/com/ibm/db2/db2jcc_license_cu/8.1/
> mkdir -p ./geronimo-1.1-SNAPSHOT/repository/com/ibm/db2/db2jcc/8.1/
> Then I copied the jars into the repo:
> cp db2jcc.jar 
> ./geronimo-1.1-SNAPSHOT/repository/com/ibm/db2/db2jcc/8.1/db2jcc-8.1.jar
> cp db2jcc_license_cu.jar 
> ./geronimo-1.1-SNAPSHOT/repository/com/ibm/db2/db2jcc_license_cu/8.1/db2jcc_license_cu-8.1.jar

> For the most part this is certainly doable but is actually pretty 
> tedious.  Also, following the convention, the artifactname and the 
> version are duplicated in the actual deployed jar; seems a bit redundant.
> I wanted to validate that my usage is correct per the original design.  
> If it is I think we're probably missing one of the points in Aaron's doc 
> about the server being simple to use.  At least this is not intuitive.
> If my understanding is wrong or there is a better suggestion on how to 
> handle this please post back.  I'm working on the documentation for how 
> to deploy DayTrader.  Since this is a very visible use case for the 
> server I want to make sure this is the direction we are going to 
> recommend for users.
> Perhaps another approach is to use the shared lib that was recently 
> added (I have to go back through the e-mails to see how to set it up).  
> I think that paradigm is more in line with Tomcat users as well as other 
> AppServers.
> Thoughts?
> Matt

View raw message