geronimo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Russell E Glaue <rgl...@cait.org>
Subject Re: Regarding GERONIMO-6317 Profiling clustering modules to save build time
Date Fri, 06 Apr 2012 16:45:53 GMT


On 04/06/2012 10:11 AM, Kevan Miller wrote:
>
> On Apr 6, 2012, at 10:47 AM, Russell E Glaue wrote:
>
>>
>> On 04/05/2012 02:49 PM, Kevan Miller wrote:
>>>
>>> On Apr 5, 2012, at 9:50 AM, Russell E Glaue wrote:
>>>
>>>> "Clustering features are not available in the current 3.0-beta branch, so
add some profiles to not build it, thus save the build time."
>>>>
>>>> Is commit 1309635 related to this?
>>>>
>>>>
>>>> Is someone able to explain why WADI Clustering and Plugin Farming are not
available in the 3.0-beta branch? Is it broken?
>>>
>>> Hi Russell,
>>> The WADI project seems to be largely dormant. Without anyone willing to bring
the function forward, I don't think anyone has looked at trying to get it to work on a 3.0
base.
>>>
>>> Is WADI important to you?
>>>
>>> --kevan
>>>
>>
>> I'm wanting to get Geronimo up in my employer's web farms, and finally getting multiple
instance support was a huge step forward for me. But now removing support for clustering is
a step backward. We had been planning on its use.
>>
>> Some mechanism is important to anyone who wants web user sessions to have persistence
among multiple clustered/loadbalanced Geronimo servers. And we have had (used to anyway) lots
of conversations on the importance of Geronimo clustering capabilities and how to do it.
>> https://cwiki.apache.org/GMOxDEV/clustering-initial-discussion.html
>>
>> Yes.., we can use other 3rd party like Terracotta, but that does not make Geronimo
enterprise-ready out-of-the-box. And the terracotta plugin is probably not updated for 3.0.
>> https://cwiki.apache.org/GMOxDEV/clustering-geronimo-with-open-terracotta.html
>>
>> If we remove WADI support, will we replace it with some other mechanism to share
web user sessions among Geronimo instances? Or will we tell users to rely on tomcat's and
jetty's native session replication and management? This alternative will not be a clean configuration,
since the changes would have manually to go into tomcat and jetty server config files - not
deployed as plans.
>
> OK. So, for session replication, most people that I know of use Tomcat native clustering
(https://cwiki.apache.org/GMOxDOC30/tomcat-native-clustering.html). Does that not work for
you? If we can go into specific issues, that would be great...
>
> Great conversation to be having. Thanks!
>
> --kevan

It is not that the tomcat alternative is not going to work for me. What I said 
about that was "This alternative will not be a clean configuration, since the 
changes would have to manually go into tomcat and jetty server config files - 
not deployed as plans."

After Forest and I got the multiple-instance support squared away for Geronimo, 
I have been working towards how to configure Geronimo by only deploying plans, 
and documenting it (I had been updating the Geronimo wiki as I go along - but 
got side tracked recently). This is currently the needed procedure for remote 
configuration.

The document, https://cwiki.apache.org/GMOxDOC30/tomcat-native-clustering.html 
(which seems to be discussing Tomcat 6) discusses how you must manually edit the 
server.xml file to "configure the engine".

There is thin if no documentation on bundling server config file in like a car 
to deploy along with a plan. But I guess we'd have to do it that way if we stick 
to the native tomcat/jetty approaches to having clustering in Geronimo.

Thoughts?

-RG

Mime
View raw message