You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@geronimo.apache.org by pdennis <pa...@managementdynamics.com> on 2008/10/15 22:38:52 UTC
Multiple Instances With One Repository
Hi,
I've looked around the documentation and the forum to find my answer and I
haven't found it, yet.
I just read about Multiple Repositories
http://cwiki.apache.org/GMOxDOC20/multiple-repositories.html here , so I can
start several instances and create a local repository for each instance.
What I want to do is to create a Repository, deploy an ear and a database
pool, and start several instances that use that Repository.
In the WASCE_HOME directory, there is the main repository and var
directories, what I would like to do is to create a local repository, and
start several instances that use that repository. So I think I am after a
structure like this:
WASCE_HOME
repo1
instance1
instance2
instance3
repo2
instance1
instance2
instance3
repo3
instance1
instance2
instance3
Is this possible? If so, how would I do this?
Thanks,
Patrick
--
View this message in context: http://www.nabble.com/Multiple-Instances-With-One-Repository-tp20001882s134p20001882.html
Sent from the Apache Geronimo - Dev mailing list archive at Nabble.com.
Re: Multiple Instances With One Repository
Posted by David Jencks <da...@yahoo.com>.
On Oct 17, 2008, at 7:02 AM, randersen wrote:
>
>
>
> djencks wrote:
>>
>>
>> On Oct 16, 2008, at 11:56 AM, randersen wrote:
>>
>>> The structure would be more like:
>>>
>>> GERONIMO_HOME
>>> customers
>>> customer1
>>> customer1-repo
>>> instance1
>>> var
>>> instance2
>>> var
>>> customer2
>>> customer2-repo
>>> instance1
>>> var
>>>
>>> This would allow the deployment of multiple customers on 1 machine
>>> that use
>>> different databases to use the same database pool without having to
>>> deploy a
>>> second repository in each instance for the database pool. Is there
>>> a system
>>> property for the repository directory like there is for the var
>>> directory?
>>
>> I'd like to know more about what you are trying to do. If each
>> customer is using the same applications and the only difference
>> between them is which database their copy of the app needs to connect
>> to then you only need one repo. The only reason I can think of to
>> have one repo for each customer is if you have apps for one customer
>> that you need to prevent being started on other customer's servers.
>>
>> In the unlikely event you are using derby or another db that stores
>> data inside geronimo's var dir then each customer will already get a
>> new copy of derby simply because they have their own var dir.
>>
>> In the more likely event you are using an external db and each
>> customer has their own copy you have at least these two options:
>>
>> 1. use a config.xml snippet with a config-substitutions variable for
>> the db name or url (and user/pw presumably) and set these in each
>> server's var/config/config-substitutions.properties
>>
>> 2. deploy a fully configured db pool on each customer's server
>> pointing to the appropriate db and use artifact-aliases to have it
>> replace a standard or dummy db pool.
>>
>> still trying to understand what you are trying to do....
>> thanks
>> david jencks
>>
>> The only difference is the database to which they are connecting to
>> and it
>> is an external one. We will be running multiple customers (each
>> customer
>> having a different database connection) with multiple instances for
>> that
>> customer on the same host and using the same version of the
>> application.
>> The only thing that needs to change is the db connection
>> information for
>> the database pool.
>>
>> 1. Would that be this line:
>> <module name="console.dbpool/CustomerDBPool/1.0/rar"/>
>>
>> how do you specify the attributes for it?
>>
>> 2. I don't understand this part, are you saying that for each
>> customer we
>> have a separate installation of geronimo? So if I have 3 customers
>> on one
>> host I'll have 3 installed versions of the software with the instance
>> config under each one?
>>
>> I like the sound of #1, that would mean that we could have 1
>> installation
>> of geronimo and have all the instances underneath and just change
>> config.xml for the ones that would use a different db connection.
>>
>>
Both solutions involve only one installation of geronimo. Both
involve deploying the datasource as a geronimo plugin and using some
plugin features. If you have any maven familiarity at all it's
easiest to generate the plugins using maven and the car-maven-plugin,
although it's not essential. In either case I'd generate the initial
plan for the datasource using the admin console.
There are some general instructions on setting up maven plugin
projects here: http://cwiki.apache.org/GMOxDOC22/constructing-a-special-purpose-server-using-maven.html
in either case you'd create the multiple geronimo instances with
something like
mkdir -p customer1/instance1
cp -r var customer1/instance1
... repeat for all customers, instances
for (1) you probably want to do this after deploying the app and
datasource plugin and for (2) before deploying the app and datasource
plugin.
(1). You'll need the full abstract name of the
JCAManagedConnectionFactoryWrapper that represents the datasource. I
think you can determine it by turning logging to debug and starting
the server after the datasource is installed. Look at the log..... it
may take some searching.
in the car-maven-plugin configuration in the pom.xml set up a
config.xml snippet like....
<instance>
<plugin-artifact>
<config-xml-content>
<gbean name="---(name of
JCAManagedConnectionFactoryWrapper)---">
<attribute name="Database">$
{DBName}</attribute>
</gbean>
</config-xml-content>
<config-substitution
key="DBName">Customer1DB</config-substitution>
</plugin-artifact>
</instance>
The actual attribute(s) you use will depend on which tranql wrapper
you are using: any config-property name can be used as an attribute.
What this does is expose these attributes as properties in var/config/
config-substitutions.properties. After you copy the var directory
into the multiple locations you can edit each customer${x}/instance$
[y}/var/config/config-substitutions.properties to have the correct
value for that instance.
(2) Create a datasource plugin for each customer db, set up to
connect to that db. Instead of modifying the config-xml-content and
using a substitution variable you'll have each plugin replace a
default one. For the non-default datasources include something like
<instance>
<plugin-artifact>
<artifact-alias key="org.myco/
defaultDatasource//car">org.myco/datasourceX/${version}/car</artifact-
alias>
<artifact-alias key="org.myco/
defaultDatasource/${version}/car">oorg.myco/datasourceX/${version}/
car</artifact-alias>
</plugin-artifact>
</instance>
set up all the instances and deploy the appropriate datasource and the
app on each of them. The code will only get installed once into the
single repository but each installation will be using the datasource
you deployed on it. It might seem a bit mysterious but the individual
var/config/config.xml files control which plugins are started on each
server instance so by deploying the appropriate datasource plugin on
each server instance only that datasource will be started.
Hope this helps.... seems like a good topic for a help page, perhaps
I'll be able to scrape together more time to set up an example.
thanks
david jencks
>
> --
> View this message in context: http://www.nabble.com/Multiple-Instances-With-One-Repository-tp20001882s134p20033920.html
> Sent from the Apache Geronimo - Dev mailing list archive at
> Nabble.com.
>
Re: Multiple Instances With One Repository
Posted by randersen <ro...@managementdynamics.com>.
djencks wrote:
>
>
> On Oct 16, 2008, at 11:56 AM, randersen wrote:
>
>> The structure would be more like:
>>
>> GERONIMO_HOME
>> customers
>> customer1
>> customer1-repo
>> instance1
>> var
>> instance2
>> var
>> customer2
>> customer2-repo
>> instance1
>> var
>>
>> This would allow the deployment of multiple customers on 1 machine
>> that use
>> different databases to use the same database pool without having to
>> deploy a
>> second repository in each instance for the database pool. Is there
>> a system
>> property for the repository directory like there is for the var
>> directory?
>
> I'd like to know more about what you are trying to do. If each
> customer is using the same applications and the only difference
> between them is which database their copy of the app needs to connect
> to then you only need one repo. The only reason I can think of to
> have one repo for each customer is if you have apps for one customer
> that you need to prevent being started on other customer's servers.
>
> In the unlikely event you are using derby or another db that stores
> data inside geronimo's var dir then each customer will already get a
> new copy of derby simply because they have their own var dir.
>
> In the more likely event you are using an external db and each
> customer has their own copy you have at least these two options:
>
> 1. use a config.xml snippet with a config-substitutions variable for
> the db name or url (and user/pw presumably) and set these in each
> server's var/config/config-substitutions.properties
>
> 2. deploy a fully configured db pool on each customer's server
> pointing to the appropriate db and use artifact-aliases to have it
> replace a standard or dummy db pool.
>
> still trying to understand what you are trying to do....
> thanks
> david jencks
>
> The only difference is the database to which they are connecting to and it
> is an external one. We will be running multiple customers (each customer
> having a different database connection) with multiple instances for that
> customer on the same host and using the same version of the application.
> The only thing that needs to change is the db connection information for
> the database pool.
>
> 1. Would that be this line:
> <module name="console.dbpool/CustomerDBPool/1.0/rar"/>
>
> how do you specify the attributes for it?
>
> 2. I don't understand this part, are you saying that for each customer we
> have a separate installation of geronimo? So if I have 3 customers on one
> host I'll have 3 installed versions of the software with the instance
> config under each one?
>
> I like the sound of #1, that would mean that we could have 1 installation
> of geronimo and have all the instances underneath and just change
> config.xml for the ones that would use a different db connection.
>
>
--
View this message in context: http://www.nabble.com/Multiple-Instances-With-One-Repository-tp20001882s134p20033920.html
Sent from the Apache Geronimo - Dev mailing list archive at Nabble.com.
Re: Multiple Instances With One Repository
Posted by David Jencks <da...@yahoo.com>.
On Oct 16, 2008, at 11:56 AM, randersen wrote:
>
>
> djencks wrote:
>>
>>
>>> In the WASCE_HOME directory, there is the main repository and var
>>> directories, what I would like to do is to create a local
>>> repository, and
>>> start several instances that use that repository. So I think I am
>>> after a
>>> structure like this:
>>>
>>> WASCE_HOME
>>> repo1
>>> instance1
>>> instance2
>>> instance3
>>> repo2
>>> instance1
>>> instance2
>>> instance3
>>> repo3
>>> instance1
>>> instance2
>>> instance3
>>>
>>> Is this possible? If so, how would I do this?
>>
>> I'm not sure why you want mutliple repositories. I think you can do
>> what you want with only the default repository.
>>
>> We actually have sort of a sample of multiple servers sharing a
>> repository. One version that may be usable is at
>> http://people.apache.org/~djencks/failover2.tar.gz
>> To build it yourself build serve/trunk, then checkout
>>
>> svn co https://svn.apache.org/repos/asf/geronimo/sandbox/failover
>>
>> and build the pieces independently.
>>
>> There's a linux script that makes a bunch of server copies (copying
>> the var directory) and a couple scripts that start the servers on
>> different ports. This sample works on plugins so you can see plugins
>> installed on each of the "servers". The first time a plugin is
>> installed it actually gets into the repository and the installation
>> process may unpack stuff into that server's var dir. When the plugin
>> is installed on the other servers sharing the repo the only thing
>> that
>> happens is the unpacking into the var dir.
>>
>> You can do something similar with apps that you aren't thinking of as
>> plugins by deploying the app into the appropriate first server, then
>> starting the app in the other servers you want it running in.
>>
>> Productive use of this feature is just starting so please let us know
>> what is unclear or if you run into problems or have more questions.
>>
>> many thanks
>> david jencks
>>
>>
>
> The structure would be more like:
>
> WASCE_HOME
> customers
> customer1
> customer1-repo
> instance1
> var
> instance2
> var
> customer2
> customer2-repo
> instance1
> var
>
> This would allow the deployment of multiple customers on 1 machine
> that use
> different databases to use the same database pool without having to
> deploy a
> second repository in each instance for the database pool. Is there
> a system
> property for the repository directory like there is for the var
> directory?
I'd like to know more about what you are trying to do. If each
customer is using the same applications and the only difference
between them is which database their copy of the app needs to connect
to then you only need one repo. The only reason I can think of to
have one repo for each customer is if you have apps for one customer
that you need to prevent being started on other customer's servers.
In the unlikely event you are using derby or another db that stores
data inside geronimo's var dir then each customer will already get a
new copy of derby simply because they have their own var dir.
In the more likely event you are using an external db and each
customer has their own copy you have at least these two options:
1. use a config.xml snippet with a config-substitutions variable for
the db name or url (and user/pw presumably) and set these in each
server's var/config/config-substitutions.properties
2. deploy a fully configured db pool on each customer's server
pointing to the appropriate db and use artifact-aliases to have it
replace a standard or dummy db pool.
still trying to understand what you are trying to do....
thanks
david jencks
>
> --
> View this message in context: http://www.nabble.com/Multiple-Instances-With-One-Repository-tp20001882s134p20016177.html
> Sent from the Apache Geronimo - Dev mailing list archive at
> Nabble.com.
>
Re: Multiple Instances With One Repository
Posted by randersen <ro...@managementdynamics.com>.
djencks wrote:
>
>
>> In the WASCE_HOME directory, there is the main repository and var
>> directories, what I would like to do is to create a local
>> repository, and
>> start several instances that use that repository. So I think I am
>> after a
>> structure like this:
>>
>> WASCE_HOME
>> repo1
>> instance1
>> instance2
>> instance3
>> repo2
>> instance1
>> instance2
>> instance3
>> repo3
>> instance1
>> instance2
>> instance3
>>
>> Is this possible? If so, how would I do this?
>
> I'm not sure why you want mutliple repositories. I think you can do
> what you want with only the default repository.
>
> We actually have sort of a sample of multiple servers sharing a
> repository. One version that may be usable is at
> http://people.apache.org/~djencks/failover2.tar.gz
> To build it yourself build serve/trunk, then checkout
>
> svn co https://svn.apache.org/repos/asf/geronimo/sandbox/failover
>
> and build the pieces independently.
>
> There's a linux script that makes a bunch of server copies (copying
> the var directory) and a couple scripts that start the servers on
> different ports. This sample works on plugins so you can see plugins
> installed on each of the "servers". The first time a plugin is
> installed it actually gets into the repository and the installation
> process may unpack stuff into that server's var dir. When the plugin
> is installed on the other servers sharing the repo the only thing that
> happens is the unpacking into the var dir.
>
> You can do something similar with apps that you aren't thinking of as
> plugins by deploying the app into the appropriate first server, then
> starting the app in the other servers you want it running in.
>
> Productive use of this feature is just starting so please let us know
> what is unclear or if you run into problems or have more questions.
>
> many thanks
> david jencks
>
>
The structure would be more like:
WASCE_HOME
customers
customer1
customer1-repo
instance1
var
instance2
var
customer2
customer2-repo
instance1
var
This would allow the deployment of multiple customers on 1 machine that use
different databases to use the same database pool without having to deploy a
second repository in each instance for the database pool. Is there a system
property for the repository directory like there is for the var directory?
--
View this message in context: http://www.nabble.com/Multiple-Instances-With-One-Repository-tp20001882s134p20016177.html
Sent from the Apache Geronimo - Dev mailing list archive at Nabble.com.
Re: Multiple Instances With One Repository
Posted by David Jencks <da...@yahoo.com>.
On Oct 15, 2008, at 1:38 PM, pdennis wrote:
>
> Hi,
>
> I've looked around the documentation and the forum to find my answer
> and I
> haven't found it, yet.
>
> I just read about Multiple Repositories
> http://cwiki.apache.org/GMOxDOC20/multiple-repositories.html here ,
> so I can
> start several instances and create a local repository for each
> instance.
> What I want to do is to create a Repository, deploy an ear and a
> database
> pool, and start several instances that use that Repository.
>
> In the WASCE_HOME directory, there is the main repository and var
> directories, what I would like to do is to create a local
> repository, and
> start several instances that use that repository. So I think I am
> after a
> structure like this:
>
> WASCE_HOME
> repo1
> instance1
> instance2
> instance3
> repo2
> instance1
> instance2
> instance3
> repo3
> instance1
> instance2
> instance3
>
> Is this possible? If so, how would I do this?
I'm not sure why you want mutliple repositories. I think you can do
what you want with only the default repository.
We actually have sort of a sample of multiple servers sharing a
repository. One version that may be usable is at http://people.apache.org/~djencks/failover2.tar.gz
To build it yourself build serve/trunk, then checkout
svn co https://svn.apache.org/repos/asf/geronimo/sandbox/failover
and build the pieces independently.
There's a linux script that makes a bunch of server copies (copying
the var directory) and a couple scripts that start the servers on
different ports. This sample works on plugins so you can see plugins
installed on each of the "servers". The first time a plugin is
installed it actually gets into the repository and the installation
process may unpack stuff into that server's var dir. When the plugin
is installed on the other servers sharing the repo the only thing that
happens is the unpacking into the var dir.
You can do something similar with apps that you aren't thinking of as
plugins by deploying the app into the appropriate first server, then
starting the app in the other servers you want it running in.
Productive use of this feature is just starting so please let us know
what is unclear or if you run into problems or have more questions.
many thanks
david jencks
>
>
> Thanks,
>
> Patrick
> --
> View this message in context: http://www.nabble.com/Multiple-Instances-With-One-Repository-tp20001882s134p20001882.html
> Sent from the Apache Geronimo - Dev mailing list archive at
> Nabble.com.
>