You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@geronimo.apache.org by Aaron Mulder <am...@alumni.princeton.edu> on 2006/05/18 22:53:49 UTC

Multiple Server Configurations

All,

David Jencks just backported a feature that lets you create multiple
server configurations inside a single Geronimo installation.  This
affects the contents of the var/ directory, if I understand it right.
So essentially, you could create a structure like this:

geronimo/var/...   (default configuration)
geronimo/server1/var/...   ("server1" configuration)
geronimo/another/var/...   ("another" configuration)

In other words, you can create subdirectories with their own copies of
var/* and then tell Geronimo during startup to read from foo/var/*
instead of var/* using a command-line parameter.

I'd like to propose one change to this, and that is, that we eliminate
the "var" directory and set it up one of these two ways -- the
difference being whether the default server configuration is named
something like "default" or named "var":

Option 1: default configuration named "var":
geronimo/var/...   (default configuration)
geronimo/server1/...   ("server1" configuration)
geronimo/another/...   ("another" configuration)

Option 2: default configuration named e.g. "default":
geronimo/default/...   (default configuration)
geronimo/server1/...   ("server1" configuration)
geronimo/another/...   ("another" configuration)

It seems somewhat more usable to me if, for example, the log directory
is immediately underneath the server configuration directory.  For
anyone who's not real UNIX-oriented, I think it will be much nicer to
look in the configuration directory and see config/ log/ security/ etc
instead of just seeing "var".

Any thoughts on this?

Thanks,
    Aaron

Re: Multiple Server Configurations

Posted by Hiram Chirino <hi...@hiramchirino.com>.
+1

On 5/18/06, Matt Hogstrom <matt@hog�����~~����E�X�\DZj�QPC�[��n�Ki�7
��B�%¨ߨU��&��[�_Ec�S;@_�p�A�C[����!���[u.���
���$��ol>��;��\�;�ɰ0�(قh�,��VJI�w���e�MF���������ҩ����1�Gi��f��LTʹ��3R��
��
�,v��	;}틗c��t��1:����u5�m��ȹ�k~�~���Y�������9�ECGw�(�"r�?t�	�V1�"�>B�b���^d�����|��c�"�Y�
�A�нOT7{v7L����o8X���Ԋb�]Ф�刚s`
����s�J;T����Ě+�X+������+���+V�,��>V�>sr@��FZm�K��R�gG(q<)"ڑk�*t��*S��õ�8?kH���E!<X������I��rpi(_
=l��Hs���g�VC�9~Z��-O�<f�<Y�Ju�]Z׏2<�Xu�Ѵ��A�U�x)��w�OZ�	�IZ�6ӻ���J{�ZR�O����m�Q�]�?��Eq�q0]oЍ\dBt�"a�h�|T�<T�g6.�됵�a���L�Q��žę��r��"�7�<]��"R��S9��^g(:�*���+(u[��|�|{����'���9
q�Bs��(@5���X��&�9�΃�jC�S��;���gYt�#�:�K���
�3����߁���xV�k�X�V!�q���d1?8r���Nz1
-��N�ӏEz��n��k YF���^�.
M���c�9-B��	=}�NOa$i�Y8{8m~�m�Z��V
��M>+8ه8�$����(ģ�%��ܶ(�/�1�po�V���7�p��CWT�㊞�y�N��Wݦ���r��
B#2�A��S����
�asmoR�zJ�Ԗ����s��
���o�)9F��æ#���٨��5zf!0Sq��?�Ke�A��H31kB����0Z.�l�ʵ��k%�p]T�4��s�@Q�*�1a��᫣��JppA<�⭫.�]
!�F�s��3��-��_L����Ң�!�pI�\TD��2pb\^\����t������h)k�ԭ�˾���+����;��d�ŏw����%�e���r傽�<���!yɁ�i�1/aN����)�\ؓ�$��ƀ��hákF�����lv��ؗ
MC�`&�$f�X��1�n�wt�)�(F&��ͽ�b"TJK/Y!��qp���%�;I�+w_�A���ֻ�V=dؗ����,�;as��S_8
�S$��"5oGdw�
��@W��h��{X�պ��)�:�ލ��N��&��tt0a��e������\7��rs�!����a����Ѱ%B�V�RV�Q{.F�oߘc~�j�+m��Z}4�Ơz}D��/"�$S��5Jo/var/*
> > instead of var/* using a command-line parameter.
> >
> > I'd like to propose one change to this, and that is, that we eliminate
> > the "var" directory and set it up one of these two ways -- the
> > difference being whether the default server configuration is named
> > something like "default" or named "var":
> >
> > Option 1: default configuration named "var":
> > geronimo/var/...   (default configuration)
> > geronimo/server1/...   ("server1" configuration)
> > geronimo/another/...   ("another" configuration)
> >
> > Option 2: default configuration named e.g. "default":
> > geronimo/default/...   (default configuration)
> > geronimo/server1/...   ("server1" configuration)
> > geronimo/another/...   ("another" configuration)
> >
> > It seems somewhat more usable to me if, for example, the log directory
> > is immediately underneath the server configuration directory.  For
> > anyone who's not real UNIX-oriented, I think it will be much nicer to
> > look in the configuration directory and see config/ log/ security/ etc
> > instead of just seeing "var".
> >
> > Any thoughts on this?
> >
> > Thanks,
> >    Aaron
> >
> >
> >
>


-- 
Regards,
Hiram

Re: Multiple Server Configurations

Posted by Dain Sundstrom <da...@iq80.com>.
+1

-dain

On May 22, 2006, at 11:08 PM, John Sisson wrote:

> I was planning on fixing the geronimo.bat and geronimo.sh scripts  
> (as part of GERONIMO-1638) to remove the GERONIMO_BASE environment  
> variable (that is no longer valid and AFAIK never worked properly)  
> and add  GERONIMO_SERVER_NAME and GERONIMO_SERVER_DIR environment  
> variables to be in sync with the changes.
>
> If we feel that this area is subject to change (since you indicated  
> it is in the 1.1 code to play with) then maybe the following should  
> be done:
> * Remove the GERONIMO_BASE from the startup scripts to avoid confusion
> * Rename the newly introduced system properties to have have an "X"  
> prefix to indicate they are experimental (following what Dain has  
> done for other experimental properties), specifically rename:
>
> org.apache.geronimo.server.dir ---> Xorg.apache.geronimo.server.dir
> org.apache.geronimo.server.name ---> Xorg.apache.geronimo.server.name
>
> I also agree with Aaron's comment "I think it will be much nicer to  
> look in the configuration directory and see config/ log/ security/ etc
> instead of just seeing "var"".
>
> Comments?
>
> John
>
> Matt Hogstrom wrote:
>> I'm fine for leaving the code in 1.1 to play with but I'm -1 in  
>> promoting it in 1.1.  This should be a 1.2 item.  One of the  
>> reasons we end up making disruptive changes later in the release  
>> is we don't have time to think this through and we'll be unhappy  
>> with the answer and end up tweaking it next time.
>>
>> That said, for 1.2 this is really needed as clustering will  
>> probably take shape.  I'd prefer to start the discussion now and  
>> finish it in 1.2.  Here's my 2c.
>>
>> geronimo/servers/default
>> geronimo/servers/foo
>> geronimo/servers/server1
>> geronimo/servers/server2
>>
>> A major grouping off of Geronimo makes sense so we can group  
>> servers together.  It would make sense to me to leave geronimo/var  
>> as the legacy, single server and the above as the clustered  
>> convention.
>>
>> Aaron Mulder wrote:
>>> All,
>>>
>>> David Jencks just backported a feature that lets you create multiple
>>> server configurations inside a single Geronimo installation.  This
>>> affects the contents of the var/ directory, if I understand it  
>>> right.
>>> So essentially, you could create a structure like this:
>>>
>>> geronimo/var/...   (default configuration)
>>> geronimo/server1/var/...   ("server1" configuration)
>>> geronimo/another/var/...   ("another" configuration)
>>>
>>> In other words, you can create subdirectories with their own  
>>> copies of
>>> var/* and then tell Geronimo during startup to read from foo/var/*
>>> instead of var/* using a command-line parameter.
>>>
>>> I'd like to propose one change to this, and that is, that we  
>>> eliminate
>>> the "var" directory and set it up one of these two ways -- the
>>> difference being whether the default server configuration is named
>>> something like "default" or named "var":
>>>
>>> Option 1: default configuration named "var":
>>> geronimo/var/...   (default configuration)
>>> geronimo/server1/...   ("server1" configuration)
>>> geronimo/another/...   ("another" configuration)
>>>
>>> Option 2: default configuration named e.g. "default":
>>> geronimo/default/...   (default configuration)
>>> geronimo/server1/...   ("server1" configuration)
>>> geronimo/another/...   ("another" configuration)
>>>
>>> It seems somewhat more usable to me if, for example, the log  
>>> directory
>>> is immediately underneath the server configuration directory.  For
>>> anyone who's not real UNIX-oriented, I think it will be much  
>>> nicer to
>>> look in the configuration directory and see config/ log/  
>>> security/ etc
>>> instead of just seeing "var".
>>>
>>> Any thoughts on this?
>>>
>>> Thanks,
>>>    Aaron
>>>
>>>
>>>
>>


Re: Multiple Server Configurations

Posted by David Jencks <da...@yahoo.com>.
+1

david jencks

On May 22, 2006, at 11:08 PM, John Sisson wrote:

> I was planning on fixing the geronimo.bat and geronimo.sh scripts  
> (as part of GERONIMO-1638) to remove the GERONIMO_BASE environment  
> variable (that is no longer valid and AFAIK never worked properly)  
> and add  GERONIMO_SERVER_NAME and GERONIMO_SERVER_DIR environment  
> variables to be in sync with the changes.
>
> If we feel that this area is subject to change (since you indicated  
> it is in the 1.1 code to play with) then maybe the following should  
> be done:
> * Remove the GERONIMO_BASE from the startup scripts to avoid confusion
> * Rename the newly introduced system properties to have have an "X"  
> prefix to indicate they are experimental (following what Dain has  
> done for other experimental properties), specifically rename:
>
> org.apache.geronimo.server.dir ---> Xorg.apache.geronimo.server.dir
> org.apache.geronimo.server.name ---> Xorg.apache.geronimo.server.name
>
> I also agree with Aaron's comment "I think it will be much nicer to  
> look in the configuration directory and see config/ log/ security/ etc
> instead of just seeing "var"".
>
> Comments?
>
> John
>
> Matt Hogstrom wrote:
>> I'm fine for leaving the code in 1.1 to play with but I'm -1 in  
>> promoting it in 1.1.  This should be a 1.2 item.  One of the  
>> reasons we end up making disruptive changes later in the release  
>> is we don't have time to think this through and we'll be unhappy  
>> with the answer and end up tweaking it next time.
>>
>> That said, for 1.2 this is really needed as clustering will  
>> probably take shape.  I'd prefer to start the discussion now and  
>> finish it in 1.2.  Here's my 2c.
>>
>> geronimo/servers/default
>> geronimo/servers/foo
>> geronimo/servers/server1
>> geronimo/servers/server2
>>
>> A major grouping off of Geronimo makes sense so we can group  
>> servers together.  It would make sense to me to leave geronimo/var  
>> as the legacy, single server and the above as the clustered  
>> convention.
>>
>> Aaron Mulder wrote:
>>> All,
>>>
>>> David Jencks just backported a feature that lets you create multiple
>>> server configurations inside a single Geronimo installation.  This
>>> affects the contents of the var/ directory, if I understand it  
>>> right.
>>> So essentially, you could create a structure like this:
>>>
>>> geronimo/var/...   (default configuration)
>>> geronimo/server1/var/...   ("server1" configuration)
>>> geronimo/another/var/...   ("another" configuration)
>>>
>>> In other words, you can create subdirectories with their own  
>>> copies of
>>> var/* and then tell Geronimo during startup to read from foo/var/*
>>> instead of var/* using a command-line parameter.
>>>
>>> I'd like to propose one change to this, and that is, that we  
>>> eliminate
>>> the "var" directory and set it up one of these two ways -- the
>>> difference being whether the default server configuration is named
>>> something like "default" or named "var":
>>>
>>> Option 1: default configuration named "var":
>>> geronimo/var/...   (default configuration)
>>> geronimo/server1/...   ("server1" configuration)
>>> geronimo/another/...   ("another" configuration)
>>>
>>> Option 2: default configuration named e.g. "default":
>>> geronimo/default/...   (default configuration)
>>> geronimo/server1/...   ("server1" configuration)
>>> geronimo/another/...   ("another" configuration)
>>>
>>> It seems somewhat more usable to me if, for example, the log  
>>> directory
>>> is immediately underneath the server configuration directory.  For
>>> anyone who's not real UNIX-oriented, I think it will be much  
>>> nicer to
>>> look in the configuration directory and see config/ log/  
>>> security/ etc
>>> instead of just seeing "var".
>>>
>>> Any thoughts on this?
>>>
>>> Thanks,
>>>    Aaron
>>>
>>>
>>>
>>
>


Re: Multiple Server Configurations

Posted by Matt Hogstrom <ma...@hogstrom.org>.
I agree with your comments on the addition of the X so please go ahead and make the changes you 
indicated below.  Let's experiment with this and get it working well.  User feedback would be good 
for this as well.

John Sisson wrote:
> I was planning on fixing the geronimo.bat and geronimo.sh scripts (as 
> part of GERONIMO-1638) to remove the GERONIMO_BASE environment variable 
> (that is no longer valid and AFAIK never worked properly) and add  
> GERONIMO_SERVER_NAME and GERONIMO_SERVER_DIR environment variables to be 
> in sync with the changes.
> 
> If we feel that this area is subject to change (since you indicated it 
> is in the 1.1 code to play with) then maybe the following should be done:
> * Remove the GERONIMO_BASE from the startup scripts to avoid confusion
> * Rename the newly introduced system properties to have have an "X" 
> prefix to indicate they are experimental (following what Dain has done 
> for other experimental properties), specifically rename:
> 
> org.apache.geronimo.server.dir ---> Xorg.apache.geronimo.server.dir
> org.apache.geronimo.server.name ---> Xorg.apache.geronimo.server.name
> 
> I also agree with Aaron's comment "I think it will be much nicer to look 
> in the configuration directory and see config/ log/ security/ etc
> instead of just seeing "var"".
> 
> Comments?
> 
> John
> 
> Matt Hogstrom wrote:
>> I'm fine for leaving the code in 1.1 to play with but I'm -1 in 
>> promoting it in 1.1.  This should be a 1.2 item.  One of the reasons 
>> we end up making disruptive changes later in the release is we don't 
>> have time to think this through and we'll be unhappy with the answer 
>> and end up tweaking it next time.
>>
>> That said, for 1.2 this is really needed as clustering will probably 
>> take shape.  I'd prefer to start the discussion now and finish it in 
>> 1.2.  Here's my 2c.
>>
>> geronimo/servers/default
>> geronimo/servers/foo
>> geronimo/servers/server1
>> geronimo/servers/server2
>>
>> A major grouping off of Geronimo makes sense so we can group servers 
>> together.  It would make sense to me to leave geronimo/var as the 
>> legacy, single server and the above as the clustered convention.
>>
>> Aaron Mulder wrote:
>>> All,
>>>
>>> David Jencks just backported a feature that lets you create multiple
>>> server configurations inside a single Geronimo installation.  This
>>> affects the contents of the var/ directory, if I understand it right.
>>> So essentially, you could create a structure like this:
>>>
>>> geronimo/var/...   (default configuration)
>>> geronimo/server1/var/...   ("server1" configuration)
>>> geronimo/another/var/...   ("another" configuration)
>>>
>>> In other words, you can create subdirectories with their own copies of
>>> var/* and then tell Geronimo during startup to read from foo/var/*
>>> instead of var/* using a command-line parameter.
>>>
>>> I'd like to propose one change to this, and that is, that we eliminate
>>> the "var" directory and set it up one of these two ways -- the
>>> difference being whether the default server configuration is named
>>> something like "default" or named "var":
>>>
>>> Option 1: default configuration named "var":
>>> geronimo/var/...   (default configuration)
>>> geronimo/server1/...   ("server1" configuration)
>>> geronimo/another/...   ("another" configuration)
>>>
>>> Option 2: default configuration named e.g. "default":
>>> geronimo/default/...   (default configuration)
>>> geronimo/server1/...   ("server1" configuration)
>>> geronimo/another/...   ("another" configuration)
>>>
>>> It seems somewhat more usable to me if, for example, the log directory
>>> is immediately underneath the server configuration directory.  For
>>> anyone who's not real UNIX-oriented, I think it will be much nicer to
>>> look in the configuration directory and see config/ log/ security/ etc
>>> instead of just seeing "var".
>>>
>>> Any thoughts on this?
>>>
>>> Thanks,
>>>    Aaron
>>>
>>>
>>>
>>
> 
> 
> 
> 

Re: Multiple Server Configurations

Posted by John Sisson <jr...@gmail.com>.
I was planning on fixing the geronimo.bat and geronimo.sh scripts (as 
part of GERONIMO-1638) to remove the GERONIMO_BASE environment variable 
(that is no longer valid and AFAIK never worked properly) and add  
GERONIMO_SERVER_NAME and GERONIMO_SERVER_DIR environment variables to be 
in sync with the changes.

If we feel that this area is subject to change (since you indicated it 
is in the 1.1 code to play with) then maybe the following should be done:
* Remove the GERONIMO_BASE from the startup scripts to avoid confusion
* Rename the newly introduced system properties to have have an "X" 
prefix to indicate they are experimental (following what Dain has done 
for other experimental properties), specifically rename:

org.apache.geronimo.server.dir ---> Xorg.apache.geronimo.server.dir
org.apache.geronimo.server.name ---> Xorg.apache.geronimo.server.name

I also agree with Aaron's comment "I think it will be much nicer to look 
in the configuration directory and see config/ log/ security/ etc
instead of just seeing "var"".

Comments?

John

Matt Hogstrom wrote:
> I'm fine for leaving the code in 1.1 to play with but I'm -1 in 
> promoting it in 1.1.  This should be a 1.2 item.  One of the reasons 
> we end up making disruptive changes later in the release is we don't 
> have time to think this through and we'll be unhappy with the answer 
> and end up tweaking it next time.
>
> That said, for 1.2 this is really needed as clustering will probably 
> take shape.  I'd prefer to start the discussion now and finish it in 
> 1.2.  Here's my 2c.
>
> geronimo/servers/default
> geronimo/servers/foo
> geronimo/servers/server1
> geronimo/servers/server2
>
> A major grouping off of Geronimo makes sense so we can group servers 
> together.  It would make sense to me to leave geronimo/var as the 
> legacy, single server and the above as the clustered convention.
>
> Aaron Mulder wrote:
>> All,
>>
>> David Jencks just backported a feature that lets you create multiple
>> server configurations inside a single Geronimo installation.  This
>> affects the contents of the var/ directory, if I understand it right.
>> So essentially, you could create a structure like this:
>>
>> geronimo/var/...   (default configuration)
>> geronimo/server1/var/...   ("server1" configuration)
>> geronimo/another/var/...   ("another" configuration)
>>
>> In other words, you can create subdirectories with their own copies of
>> var/* and then tell Geronimo during startup to read from foo/var/*
>> instead of var/* using a command-line parameter.
>>
>> I'd like to propose one change to this, and that is, that we eliminate
>> the "var" directory and set it up one of these two ways -- the
>> difference being whether the default server configuration is named
>> something like "default" or named "var":
>>
>> Option 1: default configuration named "var":
>> geronimo/var/...   (default configuration)
>> geronimo/server1/...   ("server1" configuration)
>> geronimo/another/...   ("another" configuration)
>>
>> Option 2: default configuration named e.g. "default":
>> geronimo/default/...   (default configuration)
>> geronimo/server1/...   ("server1" configuration)
>> geronimo/another/...   ("another" configuration)
>>
>> It seems somewhat more usable to me if, for example, the log directory
>> is immediately underneath the server configuration directory.  For
>> anyone who's not real UNIX-oriented, I think it will be much nicer to
>> look in the configuration directory and see config/ log/ security/ etc
>> instead of just seeing "var".
>>
>> Any thoughts on this?
>>
>> Thanks,
>>    Aaron
>>
>>
>>
>


Re: Multiple Server Configurations

Posted by Joe Bohn <jo...@earthlink.net>.
+1 to Matt's suggestion.  I know that there are already users working on 
unstable builds of 1.1.   While I think that they understand 1.1 isn't 
final yet ... we've already forced them to change a good many things. 
  I think it would be nice to avoid any more for 1.1 so I'd vote for 
option #1.

Joe


Matt Hogstrom wrote:
> I'm fine for leaving the code in 1.1 to play with but I'm -1 in 
> promoting it in 1.1.  This should be a 1.2 item.  One of the reasons we 
> end up making disruptive changes later in the release is we don't have 
> time to think this through and we'll be unhappy with the answer and end 
> up tweaking it next time.
> 
> That said, for 1.2 this is really needed as clustering will probably 
> take shape.  I'd prefer to start the discussion now and finish it in 
> 1.2.  Here's my 2c.
> 
> geronimo/servers/default
> geronimo/servers/foo
> geronimo/servers/server1
> geronimo/servers/server2
> 
> A major grouping off of Geronimo makes sense so we can group servers 
> together.  It would make sense to me to leave geronimo/var as the 
> legacy, single server and the above as the clustered convention.
> 
> Aaron Mulder wrote:
> 
>> All,
>>
>> David Jencks just backported a feature that lets you create multiple
>> server configurations inside a single Geronimo installation.  This
>> affects the contents of the var/ directory, if I understand it right.
>> So essentially, you could create a structure like this:
>>
>> geronimo/var/...   (default configuration)
>> geronimo/server1/var/...   ("server1" configuration)
>> geronimo/another/var/...   ("another" configuration)
>>
>> In other words, you can create subdirectories with their own copies of
>> var/* and then tell Geronimo during startup to read from foo/var/*
>> instead of var/* using a command-line parameter.
>>
>> I'd like to propose one change to this, and that is, that we eliminate
>> the "var" directory and set it up one of these two ways -- the
>> difference being whether the default server configuration is named
>> something like "default" or named "var":
>>
>> Option 1: default configuration named "var":
>> geronimo/var/...   (default configuration)
>> geronimo/server1/...   ("server1" configuration)
>> geronimo/another/...   ("another" configuration)
>>
>> Option 2: default configuration named e.g. "default":
>> geronimo/default/...   (default configuration)
>> geronimo/server1/...   ("server1" configuration)
>> geronimo/another/...   ("another" configuration)
>>
>> It seems somewhat more usable to me if, for example, the log directory
>> is immediately underneath the server configuration directory.  For
>> anyone who's not real UNIX-oriented, I think it will be much nicer to
>> look in the configuration directory and see config/ log/ security/ etc
>> instead of just seeing "var".
>>
>> Any thoughts on this?
>>
>> Thanks,
>>    Aaron
>>
>>
>>
> 
> 

-- 
Joe Bohn
joe.bohn at earthlink.net

"He is no fool who gives what he cannot keep, to gain what he cannot 
lose."   -- Jim Elliot

Re: Multiple Server Configurations

Posted by Paul McMahan <pa...@gmail.com>.
My 2c is that I prefer Matt's idea of grouping the server specific
directories beneath a "servers" directory.  As a new user exploring
the directory structure I would find that intuitive.  But I'm hesitant
to base my position solely on this empathetic gut feel because I'm
wondering which parts of this discussion bump up against the
clustering decisions for Geronimo 1.2, especially w.r.t. vertical
clustering.  Maybe this discussion will end up driving the clustering
decisions, but maybe vice versa.

My other 2c (making my contribution come to a grand total of 4 cents,
wow!! can I go home now?) is that tomcat and jetty have already paved
the way for at least part of this discussion.  For example, in the
tomcat 5.5.12 distribution there's a section of RUNNING.txt devoted to
configuring multiple instances of tomcat.  In the interest of making a
transition from those environments go as smoothly as possible we
should adopt any pieces of their approach that can fit.

Best wishes,
Paul

p.s. my favorite bikeshed color is Blue. No, yello... :-)


On 5/18/06, Matt Hogstrom <ma...@hogstrom.org> wrote:
> I'm fine for leaving the code in 1.1 to play with but I'm -1 in promoting it in 1.1.  This should be
> a 1.2 item.  One of the reasons we end up making disruptive changes later in the release is we don't
> have time to think this through and we'll be unhappy with the answer and end up tweaking it next time.
>
> That said, for 1.2 this is really needed as clustering will probably take shape.  I'd prefer to
> start the discussion now and finish it in 1.2.  Here's my 2c.
>
> geronimo/servers/default
> geronimo/servers/foo
> geronimo/servers/server1
> geronimo/servers/server2
>
> A major grouping off of Geronimo makes sense so we can group servers together.  It would make sense
> to me to leave geronimo/var as the legacy, single server and the above as the clustered convention.
>
> Aaron Mulder wrote:
> > All,
> >
> > David Jencks just backported a feature that lets you create multiple
> > server configurations inside a single Geronimo installation.  This
> > affects the contents of the var/ directory, if I understand it right.
> > So essentially, you could create a structure like this:
> >
> > geronimo/var/...   (default configuration)
> > geronimo/server1/var/...   ("server1" configuration)
> > geronimo/another/var/...   ("another" configuration)
> >
> > In other words, you can create subdirectories with their own copies of
> > var/* and then tell Geronimo during startup to read from foo/var/*
> > instead of var/* using a command-line parameter.
> >
> > I'd like to propose one change to this, and that is, that we eliminate
> > the "var" directory and set it up one of these two ways -- the
> > difference being whether the default server configuration is named
> > something like "default" or named "var":
> >
> > Option 1: default configuration named "var":
> > geronimo/var/...   (default configuration)
> > geronimo/server1/...   ("server1" configuration)
> > geronimo/another/...   ("another" configuration)
> >
> > Option 2: default configuration named e.g. "default":
> > geronimo/default/...   (default configuration)
> > geronimo/server1/...   ("server1" configuration)
> > geronimo/another/...   ("another" configuration)
> >
> > It seems somewhat more usable to me if, for example, the log directory
> > is immediately underneath the server configuration directory.  For
> > anyone who's not real UNIX-oriented, I think it will be much nicer to
> > look in the configuration directory and see config/ log/ security/ etc
> > instead of just seeing "var".
> >
> > Any thoughts on this?
> >
> > Thanks,
> >    Aaron
> >
> >
> >
>

Re: Multiple Server Configurations

Posted by Matt Hogstrom <ma...@hogstrom.org>.
I'm fine for leaving the code in 1.1 to play with but I'm -1 in promoting it in 1.1.  This should be 
a 1.2 item.  One of the reasons we end up making disruptive changes later in the release is we don't 
have time to think this through and we'll be unhappy with the answer and end up tweaking it next time.

That said, for 1.2 this is really needed as clustering will probably take shape.  I'd prefer to 
start the discussion now and finish it in 1.2.  Here's my 2c.

geronimo/servers/default
geronimo/servers/foo
geronimo/servers/server1
geronimo/servers/server2

A major grouping off of Geronimo makes sense so we can group servers together.  It would make sense 
to me to leave geronimo/var as the legacy, single server and the above as the clustered convention.

Aaron Mulder wrote:
> All,
> 
> David Jencks just backported a feature that lets you create multiple
> server configurations inside a single Geronimo installation.  This
> affects the contents of the var/ directory, if I understand it right.
> So essentially, you could create a structure like this:
> 
> geronimo/var/...   (default configuration)
> geronimo/server1/var/...   ("server1" configuration)
> geronimo/another/var/...   ("another" configuration)
> 
> In other words, you can create subdirectories with their own copies of
> var/* and then tell Geronimo during startup to read from foo/var/*
> instead of var/* using a command-line parameter.
> 
> I'd like to propose one change to this, and that is, that we eliminate
> the "var" directory and set it up one of these two ways -- the
> difference being whether the default server configuration is named
> something like "default" or named "var":
> 
> Option 1: default configuration named "var":
> geronimo/var/...   (default configuration)
> geronimo/server1/...   ("server1" configuration)
> geronimo/another/...   ("another" configuration)
> 
> Option 2: default configuration named e.g. "default":
> geronimo/default/...   (default configuration)
> geronimo/server1/...   ("server1" configuration)
> geronimo/another/...   ("another" configuration)
> 
> It seems somewhat more usable to me if, for example, the log directory
> is immediately underneath the server configuration directory.  For
> anyone who's not real UNIX-oriented, I think it will be much nicer to
> look in the configuration directory and see config/ log/ security/ etc
> instead of just seeing "var".
> 
> Any thoughts on this?
> 
> Thanks,
>    Aaron
> 
> 
> 

Re: Multiple Server Configurations

Posted by David Jencks <da...@yahoo.com>.
On May 18, 2006, at 1:53 PM, Aaron Mulder wrote:

> All,
>
> David Jencks just backported a feature that lets you create multiple
> server configurations inside a single Geronimo installation.  This
> affects the contents of the var/ directory, if I understand it right.
> So essentially, you could create a structure like this:
>
> geronimo/var/...   (default configuration)
> geronimo/server1/var/...   ("server1" configuration)
> geronimo/another/var/...   ("another" configuration)
>
> In other words, you can create subdirectories with their own copies of
> var/* and then tell Geronimo during startup to read from foo/var/*
> instead of var/* using a command-line parameter.
>
> I'd like to propose one change to this, and that is, that we eliminate
> the "var" directory and set it up one of these two ways -- the
> difference being whether the default server configuration is named
> something like "default" or named "var":
>
> Option 1: default configuration named "var":
> geronimo/var/...   (default configuration)
> geronimo/server1/...   ("server1" configuration)
> geronimo/another/...   ("another" configuration)
>
> Option 2: default configuration named e.g. "default":
> geronimo/default/...   (default configuration)
> geronimo/server1/...   ("server1" configuration)
> geronimo/another/...   ("another" configuration)
>
> It seems somewhat more usable to me if, for example, the log directory
> is immediately underneath the server configuration directory.  For
> anyone who's not real UNIX-oriented, I think it will be much nicer to
> look in the configuration directory and see config/ log/ security/ etc
> instead of just seeing "var".
>
> Any thoughts on this?


Any location specified in a module other than the bootstrap module  
(typically j2ee-system) can be modified in that server's config.xml.   
If you want to modify the location of anything specified in the  
bootstrap module you will either need to modify it globally or  
provide a system property to override it.

Personally I kind of like the var, but I don't really care.  I think  
its a bikeshed and would prefer not to introduce more system properties.

thanks
david jencks

>
> Thanks,
>    Aaron


Re: Multiple Server Configurations

Posted by ia...@jpmchase.com.
I've added a new skeleton document, Geronimo v1.1 - Multiple Server
Configurations
(http://opensource.atlassian.com/confluence/oss/display/GERONIMO/Geronimo+v1.1+-+Multiple+Server+Configurations)
 to the Confluence Wiki if someone more knowledgable (*cough* Hernan
*cough*) wants to flesh out the content.


Ian

It's better to be hated for who you are
than loved for who you are not

Ian D. Stewart
Appl Dev Analyst-Advisory, DCS Automation
JPMorganChase Global Technology Infrastructure
Phone: (614) 244-2564
Pager: (888) 260-0078


                                                                                                                                          
                      "Aaron Mulder"                                                                                                      
                      <ammulder@alumni.pri        To:       "Geronimo Dev" <de...@geronimo.apache.org>                                      
                      nceton.edu>                 cc:                                                                                     
                      Sent by:                    Subject:  Multiple Server Configurations                                                
                      ammulder@gmail.com                                                                                                  
                                                                                                                                          
                                                                                                                                          
                      05/18/2006 04:53 PM                                                                                                 
                      Please respond to                                                                                                   
                      dev                                                                                                                 
                                                                                                                                          




All,

David Jencks just backported a feature that lets you create multiple
server configurations inside a single Geronimo installation.  This
affects the contents of the var/ directory, if I understand it right.
So essentially, you could create a structure like this:

geronimo/var/...   (default configuration)
geronimo/server1/var/...   ("server1" configuration)
geronimo/another/var/...   ("another" configuration)

In other words, you can create subdirectories with their own copies of
var/* and then tell Geronimo during startup to read from foo/var/*
instead of var/* using a command-line parameter.

I'd like to propose one change to this, and that is, that we eliminate
the "var" directory and set it up one of these two ways -- the
difference being whether the default server configuration is named
something like "default" or named "var":

Option 1: default configuration named "var":
geronimo/var/...   (default configuration)
geronimo/server1/...   ("server1" configuration)
geronimo/another/...   ("another" configuration)

Option 2: default configuration named e.g. "default":
geronimo/default/...   (default configuration)
geronimo/server1/...   ("server1" configuration)
geronimo/another/...   ("another" configuration)

It seems somewhat more usable to me if, for example, the log directory
is immediately underneath the server configuration directory.  For
anyone who's not real UNIX-oriented, I think it will be much nicer to
look in the configuration directory and see config/ log/ security/ etc
instead of just seeing "var".

Any thoughts on this?

Thanks,
    Aaron