Skip to main content

JCA adapter and load balancing

17 replies [Last post]
Anonymous

Hello,

I just read the great design overview of glassfish v2 on the wiki. I
might have missed something but all I could find is load balancing for
the web tier. But what about load balancing the business tier? If I need
high availability of the business tier for rich clients, how to do that?
How is the state of the stateful session beans clustered? How to integrate
JCA adapters which hold a connection to another system?

Regards
Haug

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@glassfish.dev.java.net
For additional commands, e-mail: users-help@glassfish.dev.java.net

Reply viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.
sivakumart
Offline
Joined: 2004-01-30

As mentioned by others in the thread, there exists no out-of-the-box mechanism to make an RA instance in a cluster as the active one. This could be achieved either using Shoal or the RA could use custom group management frameworks [like JGroups/JXTA] to co-ordinate the active and failover candidate RA instances in a cluster.

Thanks
--Siva.
http://blogs.sun.com/sivakumart

Shreedhar Ganapathy

glassfish@javadesktop.org wrote:
> As mentioned by others in the thread, there exists no out-of-the-box mechanism to make an RA instance in a cluster as the active one. This could be achieved either using Shoal or the RA could use custom group management frameworks [like JGroups/JXTA] to co-ordinate the active and failover candidate RA instances in a cluster.
>
That said, I should add that the benefit with using Shoal is that
a. its already available in the appserver
b. one is not bound to a specific group communication implementation
such as Jgroups or Jxta, as it provides an API that clients can talk to.
Currently our service provider is JXTA based, and if a JGroups based
service is needed, a service provider implementation will be all that
will need to be added.
> Thanks
> --Siva.
> http://blogs.sun.com/sivakumart
> [Message sent by forum member 'sivakumart' (sivakumart)]
>
> http://forums.java.net/jive/thread.jspa?messageID=204237
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@glassfish.dev.java.net
> For additional commands, e-mail: users-help@glassfish.dev.java.net
>
>

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@glassfish.dev.java.net
For additional commands, e-mail: users-help@glassfish.dev.java.net

pinus
Offline
Joined: 2005-12-21

Wow, what a thread. I do admit that I could not follow the discussion but it seems that shoal is the solution for a clustered JCA adapter with a single connection to an external system. And if others do not know what shoal is, like me, I added the link to the project.

"Shoal is a java based clustering framework that provides infrastructure to build fault tolerance, reliability and availability."
https://shoal.dev.java.net/

A guide to develop JCA adapters.
https://glassfish.dev.java.net/javaee5/docs/DG/bealk.html

bidibidi
Offline
Joined: 2008-04-17

hi guys!

long time since last post was written here, but unfortunately i ran into the same problem..
jca which holds tcp connection in clustered environment.. fortunately everything looks fine except one detail - in Shoal i'm using DistributedStateCache to store the state of active RA. i get an exception when trying to get from cache serialized class (my own!) which was put there by another instance.. everything works fine when i use String to store state, but with my custom, even very simple class:

public class CachedData implements Serializable {
String value;
private static final long serialVersionUID = 7526471155622776147L;

public void setValue(String value) {
this.value = value;
}

public String getValue() {
return value;
}
public CachedData() {
this("empty");
}
public CachedData(String val) {
this.setValue(val);
}
}

the class cannot get from cache states added by another instacnes and only
java.lang.ClassNotFoundException: procon.adapter.cluster.CachedData

is printed..
i know that during joing the group this synchronization is performed, but how to ensure that class will be found??

thanks in avance for your replies!

Sivakumar Thyagarajan

Hi

It is good to hear about the progress you have made.

>> the class cannot get from cache states added by another instacnes and only
>> java.lang.ClassNotFoundException: procon.adapter.cluster.CachedData
>>
>> is printed..
>> i know that during joing the group this synchronization is performed, but how to ensure that class will be found??

This appears to be a classloading issue (and may not be Shoal's). Could
you let us know where procon.adapter.cluster.CachedData is placed in the
classpath? Are you bundling it in the RAR? Where are the Shoal libraries
placed in the RAR?

In short, the CachedData and other similar classes should be placed in
the classloader hierarchy [1] so that it is visible to both the RAR and
Shoal. Does placing this shared library in classpath-suffix help?

Thanks
--Siva.
[1] https://glassfish.dev.java.net/nonav/javaee5/docs/DG/beade.html

glassfish@javadesktop.org wrote:
> hi guys!

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@glassfish.dev.java.net
For additional commands, e-mail: users-help@glassfish.dev.java.net

bidibidi
Offline
Joined: 2008-04-17

thanks a lot for your advices!
as you wrote the problem was with class loading. now everything works perfectly!

thanks once more for help!

pinus
Offline
Joined: 2005-12-21

If you have a JCA adapter that uses a TCP connection to an external system (EIS, host, whatever) it is easy in an unclustered environment. But if you deploy the JCA adapter into two JVMs you need to have two connections to the external system which might not be possible.
What I want is a cluster singleton. I want to deploy the same JCA adapter to all JVMs in the cluster but only one is active at a time and holds the connection to the external system. The inactive instances reroute all calls to the active adapter. If the JVM with the active adapter fails, crashes ... the inactive adapter takes over and establishes the connection to the external system. For that reason I need a cluster singleton for the TCP connection.

fkieviet
Offline
Joined: 2003-06-14

Hi,

Incidentally, I recently ran into a similar situation.

Can the mechanism that the HTTP session replication / SFSB uses, i.e. GSM, memory replication, etc. be generalized so that it can be reused in other components?

Kind regards,

Frank Kieviet
http://blogs.sun.com/fkieviet

Shreedhar Ganapathy

Do you mean Shoal GMS ?
I would need more data to talk about how JCA connectors could be clustered.
Shoal GMS is a generalized component that is independently employable
in-process in a java application or appserver or any component in a VM
to group VMs.

glassfish@javadesktop.org wrote:
> Hi,
>
> Incidentally, I recently ran into a similar situation.
>
> Can the mechanism that the HTTP session replication / SFSB uses, i.e. GSM, memory replication, etc. be generalized so that it can be reused in other components?
>
> Kind regards,
>
> Frank Kieviet
> http://blogs.sun.com/fkieviet
> [Message sent by forum member 'fkieviet' (fkieviet)]
>
> http://forums.java.net/jive/thread.jspa?messageID=203962
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@glassfish.dev.java.net
> For additional commands, e-mail: users-help@glassfish.dev.java.net
>
>
[att1.html]

Binod

In general, this is a question about how a resource adapter or
a JBI component leverage appserver's clustering capabilities.

For this discussion, you can think of the resource adapter as yet
another container like EJB container or Web container.

The question really is that, are our clustering infrastructure like
GMS and in-memory replication reusable?

thanks,
Binod.

> Do you mean Shoal GMS ?
> I would need more data to talk about how JCA connectors could be
> clustered.
> Shoal GMS is a generalized component that is independently employable
> in-process in a java application or appserver or any component in a VM
> to group VMs.
>
> glassfish@javadesktop.org wrote:
>
>>Hi,
>>
>>Incidentally, I recently ran into a similar situation.
>>
>>Can the mechanism that the HTTP session replication / SFSB uses, i.e. GSM, memory replication, etc. be generalized so that it can be reused in other components?
>>
>>Kind regards,
>>
>>Frank Kieviet
>>http://blogs.sun.com/fkieviet
>>[Message sent by forum member 'fkieviet' (fkieviet)]
>>
>>http://forums.java.net/jive/thread.jspa?messageID=203962
>>
>>---------------------------------------------------------------------
>>To unsubscribe, e-mail: users-unsubscribe@glassfish.dev.java.net
>>For additional commands, e-mail: users-help@glassfish.dev.java.net
>>
>>
>>

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@glassfish.dev.java.net
For additional commands, e-mail: users-help@glassfish.dev.java.net

Shreedhar Ganapathy

Binod wrote:
> In general, this is a question about how a resource adapter or
> a JBI component leverage appserver's clustering capabilities.
>
> For this discussion, you can think of the resource adapter as yet
> another container like EJB container or Web container.
In such a case, the RA or JCA or possibly an object representing a
collection of Adapters (lets call it a RAFacade) is simply a client
component for GMS within a clustered instance. And gets notified when
it can participate in the group of instances.And once in the group, the
RAFacade can take actions to manage its state and RAs in different
instances can talk to each other in a cluster if need be. That said,
solution for clustered instance access to a backend EIS could be in
found through more than one approach.
>
> The question really is that, are our clustering infrastructure like
> GMS and in-memory replication reusable?
GMS is certainly designed for reusability assuming that within a VM,
multiple components can use it for their own purposes in a
group/collection context.
I am not sure I understand how in-memory replication comes into the
picture here.
>
> thanks,
> Binod.
>
>> Do you mean Shoal GMS ?
>> I would need more data to talk about how JCA connectors could be
>> clustered.
>> Shoal GMS is a generalized component that is independently employable
>> in-process in a java application or appserver or any component in a
>> VM to group VMs.
>>
>> glassfish@javadesktop.org wrote:
>>
>>> Hi,
>>>
>>> Incidentally, I recently ran into a similar situation.
>>>
>>> Can the mechanism that the HTTP session replication / SFSB uses,
>>> i.e. GSM, memory replication, etc. be generalized so that it can be
>>> reused in other components?
>>>
>>> Kind regards,
>>>
>>> Frank Kieviet
>>> http://blogs.sun.com/fkieviet
>>> [Message sent by forum member 'fkieviet' (fkieviet)]
>>>
>>> http://forums.java.net/jive/thread.jspa?messageID=203962
>>>
>>> ---------------------------------------------------------------------
>>> To unsubscribe, e-mail: users-unsubscribe@glassfish.dev.java.net
>>> For additional commands, e-mail: users-help@glassfish.dev.java.net
>>>
>>>
>>>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@glassfish.dev.java.net
> For additional commands, e-mail: dev-help@glassfish.dev.java.net
>

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@glassfish.dev.java.net
For additional commands, e-mail: users-help@glassfish.dev.java.net

Sivakumar Thyagarajan

Hi Shreedhar

Thanks for the response.

Shreedhar Ganapathy wrote:
>
> Binod wrote:
>> In general, this is a question about how a resource adapter or
>> a JBI component leverage appserver's clustering capabilities.
>>
>> For this discussion, you can think of the resource adapter as yet
>> another container like EJB container or Web container.
> In such a case, the RA or JCA or possibly an object representing a
> collection of Adapters (lets call it a RAFacade) is simply a client
> component for GMS within a clustered instance. And gets notified when
> it can participate in the group of instances.And once in the group, the
> RAFacade can take actions to manage its state and RAs in different
> instances can talk to each other in a cluster if need be. That said,
> solution for clustered instance access to a backend EIS could be in
> found through more than one approach.
Do we have any samples/documentation showing how components deployed in
GlassFish [EJB/Connectors] could use GMS to co-ordinate activities within
themselves in a cluster?

>>
>> The question really is that, are our clustering infrastructure like
>> GMS and in-memory replication reusable?
> GMS is certainly designed for reusability assuming that within a VM,
> multiple components can use it for their own purposes in a
> group/collection context. I am not sure I understand how in-memory
> replication comes into the picture here.
I think they were wanting some form of in-memory replication to keep track of RA
state that could be shared across RA instances in a cluster.

Thanks
--Siva
>>
>> thanks,
>> Binod.
>>
>>> Do you mean Shoal GMS ?
>>> I would need more data to talk about how JCA connectors could be
>>> clustered.
>>> Shoal GMS is a generalized component that is independently employable
>>> in-process in a java application or appserver or any component in a
>>> VM to group VMs.
>>>
>>> glassfish@javadesktop.org wrote:
>>>
>>>> Hi,
>>>>
>>>> Incidentally, I recently ran into a similar situation.
>>>>
>>>> Can the mechanism that the HTTP session replication / SFSB uses,
>>>> i.e. GSM, memory replication, etc. be generalized so that it can be
>>>> reused in other components?
>>>>
>>>> Kind regards,
>>>>
>>>> Frank Kieviet
>>>> http://blogs.sun.com/fkieviet
>>>> [Message sent by forum member 'fkieviet' (fkieviet)]
>>>>
>>>> http://forums.java.net/jive/thread.jspa?messageID=203962
>>>>
>>>> ---------------------------------------------------------------------
>>>> To unsubscribe, e-mail: users-unsubscribe@glassfish.dev.java.net
>>>> For additional commands, e-mail: users-help@glassfish.dev.java.net
>>>>
>>>>
>>>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: dev-unsubscribe@glassfish.dev.java.net
>> For additional commands, e-mail: dev-help@glassfish.dev.java.net
>>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@glassfish.dev.java.net
> For additional commands, e-mail: dev-help@glassfish.dev.java.net
>

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@glassfish.dev.java.net
For additional commands, e-mail: users-help@glassfish.dev.java.net

Shreedhar Ganapathy

Hi Siva
Some responses below:

>> In such a case, the RA or JCA or possibly an object representing a
>> collection of Adapters (lets call it a RAFacade) is simply a client
>> component for GMS within a clustered instance. And gets notified
>> when it can participate in the group of instances.And once in the
>> group, the RAFacade can take actions to manage its state and RAs in
>> different instances can talk to each other in a cluster if need be.
>> That said, solution for clustered instance access to a backend EIS
>> could be in found through more than one approach.
> Do we have any samples/documentation showing how components deployed
> in GlassFish [EJB/Connectors] could use GMS to co-ordinate activities
> within themselves in a cluster?
In GlassFish we have components such as transaction recovery, Iiop
failover loadbalancer using Shoal. That may help. Additionally, the
Shoal example uses a mock Application server along with two service
threads named "EJBContainer" and "TransactionService" to demonstrate how
these components in an app server coordinate with each other.

Look here :
https://shoal.dev.java.net/source/browse/shoal/gms/tests/com/sun/enterpr...
more below:
>
>>>
>>> The question really is that, are our clustering infrastructure like
>>> GMS and in-memory replication reusable?
>> GMS is certainly designed for reusability assuming that within a VM,
>> multiple components can use it for their own purposes in a
>> group/collection context. I am not sure I understand how in-memory
>> replication comes into the picture here.
> I think they were wanting some form of in-memory replication to keep
> track of RA state that could be shared across RA instances in a cluster.
Does not seem like a high throughput activity. Shoal provides the
ability to send (and receive) messages to (from) target components in
target instances thus allowing an RA component to address a message to
the group( or individual member) with target component, say "RA". This
is one way to track RA state.
Another way is to use the provided default implementation of
DistributedStateCache which is a shared cache across all instances.
Putting the RA state in the DSC would make it available in all
instances. An RA manager could simply then consult the DSC, for
instance, to find out which RA is in active state.

Caveat: I am not sure if this RA caching is something we should consider
for 9.1 FCS at this late stage. We should take into account testing
resources(or lack thereof) and the riskiness of adding such features
late in the day.
>
> Thanks
> --Siva
>>>
>>> thanks,
>>> Binod.
>>>
>>>> Do you mean Shoal GMS ?
>>>> I would need more data to talk about how JCA connectors could be
>>>> clustered.
>>>> Shoal GMS is a generalized component that is independently
>>>> employable in-process in a java application or appserver or any
>>>> component in a VM to group VMs.
>>>>
>>>> glassfish@javadesktop.org wrote:
>>>>
>>>>> Hi,
>>>>>
>>>>> Incidentally, I recently ran into a similar situation.
>>>>>
>>>>> Can the mechanism that the HTTP session replication / SFSB uses,
>>>>> i.e. GSM, memory replication, etc. be generalized so that it can
>>>>> be reused in other components?
>>>>>
>>>>> Kind regards,
>>>>>
>>>>> Frank Kieviet
>>>>> http://blogs.sun.com/fkieviet
>>>>> [Message sent by forum member 'fkieviet' (fkieviet)]
>>>>>
>>>>> http://forums.java.net/jive/thread.jspa?messageID=203962
>>>>>
>>>>> ---------------------------------------------------------------------
>>>>> To unsubscribe, e-mail: users-unsubscribe@glassfish.dev.java.net
>>>>> For additional commands, e-mail: users-help@glassfish.dev.java.net
>>>>>
>>>>>
>>>>>
>>>
>>> ---------------------------------------------------------------------
>>> To unsubscribe, e-mail: dev-unsubscribe@glassfish.dev.java.net
>>> For additional commands, e-mail: dev-help@glassfish.dev.java.net
>>>
>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: dev-unsubscribe@glassfish.dev.java.net
>> For additional commands, e-mail: dev-help@glassfish.dev.java.net
>>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@glassfish.dev.java.net
> For additional commands, e-mail: dev-help@glassfish.dev.java.net
>

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@glassfish.dev.java.net
For additional commands, e-mail: users-help@glassfish.dev.java.net

Sivakumar Thyagarajan

Hi Shreedhar

Responses inline.

Shreedhar Ganapathy wrote:
> Hi Siva
> Some responses below:
>
>>> In such a case, the RA or JCA or possibly an object representing a
>>> collection of Adapters (lets call it a RAFacade) is simply a client
>>> component for GMS within a clustered instance. And gets notified
>>> when it can participate in the group of instances.And once in the
>>> group, the RAFacade can take actions to manage its state and RAs in
>>> different instances can talk to each other in a cluster if need be.
>>> That said, solution for clustered instance access to a backend EIS
>>> could be in found through more than one approach.
>> Do we have any samples/documentation showing how components deployed
>> in GlassFish [EJB/Connectors] could use GMS to co-ordinate activities
>> within themselves in a cluster?
> In GlassFish we have components such as transaction recovery, Iiop
> failover loadbalancer using Shoal. That may help. Additionally, the
> Shoal example uses a mock Application server along with two service
> threads named "EJBContainer" and "TransactionService" to demonstrate how
> these components in an app server coordinate with each other.
>
> Look here :
> https://shoal.dev.java.net/source/browse/shoal/gms/tests/com/sun/enterpr...
Thanks, this helps.
>
> more below:
>>
>>>>
>>>> The question really is that, are our clustering infrastructure like
>>>> GMS and in-memory replication reusable?
>>> GMS is certainly designed for reusability assuming that within a VM,
>>> multiple components can use it for their own purposes in a
>>> group/collection context. I am not sure I understand how in-memory
>>> replication comes into the picture here.
>> I think they were wanting some form of in-memory replication to keep
>> track of RA state that could be shared across RA instances in a cluster.
> Does not seem like a high throughput activity. Shoal provides the
> ability to send (and receive) messages to (from) target components in
> target instances thus allowing an RA component to address a message to
> the group( or individual member) with target component, say "RA". This
> is one way to track RA state.
> Another way is to use the provided default implementation of
> DistributedStateCache which is a shared cache across all instances.
> Putting the RA state in the DSC would make it available in all
> instances. An RA manager could simply then consult the DSC, for
> instance, to find out which RA is in active state.
This DistributedStateCache should work. Thanks.
>
> Caveat: I am not sure if this RA caching is something we should consider
> for 9.1 FCS at this late stage. We should take into account testing
> resources(or lack thereof) and the riskiness of adding such features
> late in the day.
I am sorry if I was not clear. This discussion was to figure out what a resource
adapter [a 3rd party RA deployed in GlassFish] could do to achieve the above
requirements. I don't think the connector container in GlassFish should provide
this feature as of now.

Thanks
--Siva.
>>
>> Thanks
>> --Siva
>>>>
>>>> thanks,
>>>> Binod.
>>>>
>>>>> Do you mean Shoal GMS ?
>>>>> I would need more data to talk about how JCA connectors could be
>>>>> clustered.
>>>>> Shoal GMS is a generalized component that is independently
>>>>> employable in-process in a java application or appserver or any
>>>>> component in a VM to group VMs.
>>>>>
>>>>> glassfish@javadesktop.org wrote:
>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> Incidentally, I recently ran into a similar situation.
>>>>>>
>>>>>> Can the mechanism that the HTTP session replication / SFSB uses,
>>>>>> i.e. GSM, memory replication, etc. be generalized so that it can
>>>>>> be reused in other components?
>>>>>>
>>>>>> Kind regards,
>>>>>>
>>>>>> Frank Kieviet
>>>>>> http://blogs.sun.com/fkieviet
>>>>>> [Message sent by forum member 'fkieviet' (fkieviet)]
>>>>>>
>>>>>> http://forums.java.net/jive/thread.jspa?messageID=203962
>>>>>>
>>>>>> ---------------------------------------------------------------------
>>>>>> To unsubscribe, e-mail: users-unsubscribe@glassfish.dev.java.net
>>>>>> For additional commands, e-mail: users-help@glassfish.dev.java.net
>>>>>>
>>>>>>
>>>>>>
>>>>
>>>> ---------------------------------------------------------------------
>>>> To unsubscribe, e-mail: dev-unsubscribe@glassfish.dev.java.net
>>>> For additional commands, e-mail: dev-help@glassfish.dev.java.net
>>>>
>>>
>>>
>>> ---------------------------------------------------------------------
>>> To unsubscribe, e-mail: dev-unsubscribe@glassfish.dev.java.net
>>> For additional commands, e-mail: dev-help@glassfish.dev.java.net
>>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: dev-unsubscribe@glassfish.dev.java.net
>> For additional commands, e-mail: dev-help@glassfish.dev.java.net
>>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@glassfish.dev.java.net
> For additional commands, e-mail: dev-help@glassfish.dev.java.net
>

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@glassfish.dev.java.net
For additional commands, e-mail: users-help@glassfish.dev.java.net

Sivakumar Thyagarajan

> Regarding JCA: Could you give more details on what you need? GlassFish
> supports standard JCA 1.5 Adapters.
Yes please provide more details on what exactly you are looking for. GlassFish's
support for Connectors is detailed at
https://glassfish.dev.java.net/javaee5/docs/DG/bealk.html

--Siva.

Daniel Adelhardt wrote:
> Hello Haug,
>
> GlassFish V2 will support Loadbalancing and Failover on EJBs as well.
>
> See the specs for GF V2 IIOP Failover:
> http://www.glassfishwiki.org/gfwiki/attach/OnePagersOrFunctionalSpecs/FO...
>
> and the in memory state replication for HTTP Sessions and statefull
> session beans:
> http://www.glassfishwiki.org/gfwiki/attach/OnePagersOrFunctionalSpecs/me...
>
>
> Regarding JCA: Could you give more details on what you need? GlassFish
> supports standard JCA 1.5 Adapters.
>
> regards
> Daniel
>
>
> Haug Bürger wrote:
>> Hello,
>>
>> I just read the great design overview of glassfish v2 on the wiki. I
>> might have missed something but all I could find is load balancing for
>> the web tier. But what about load balancing the business tier? If I need
>> high availability of the business tier for rich clients, how to do that?
>> How is the state of the stateful session beans clustered? How to
>> integrate JCA adapters which hold a connection to another system?
>>
>> Regards
>> Haug
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: users-unsubscribe@glassfish.dev.java.net
>> For additional commands, e-mail: users-help@glassfish.dev.java.net
>>
>>
>
>

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@glassfish.dev.java.net
For additional commands, e-mail: users-help@glassfish.dev.java.net

Daniel Adelhardt

Hello Haug,

GlassFish V2 will support Loadbalancing and Failover on EJBs as well.

See the specs for GF V2 IIOP Failover:
http://www.glassfishwiki.org/gfwiki/attach/OnePagersOrFunctionalSpecs/FO...
and the in memory state replication for HTTP Sessions and statefull
session beans:
http://www.glassfishwiki.org/gfwiki/attach/OnePagersOrFunctionalSpecs/me...

Regarding JCA: Could you give more details on what you need? GlassFish
supports standard JCA 1.5 Adapters.

regards
Daniel

Haug Bürger wrote:
> Hello,
>
> I just read the great design overview of glassfish v2 on the wiki. I
> might have missed something but all I could find is load balancing for
> the web tier. But what about load balancing the business tier? If I need
> high availability of the business tier for rich clients, how to do that?
> How is the state of the stateful session beans clustered? How to integrate
> JCA adapters which hold a connection to another system?
>
> Regards
> Haug
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@glassfish.dev.java.net
> For additional commands, e-mail: users-help@glassfish.dev.java.net
>
>

--
*****************************************************************
Sun Microsystems GmbH Daniel Adelhardt
Phone: (+49 89) 46008 2443
Sonnenallee 1 Mobile: (+49 172) 8417283
85551 Kirchheim-Heimstetten Fax: (+49 89) 46008 2570
http://www.sun.de mailto:dadelhardt@sun.com
*****************************************************************

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@glassfish.dev.java.net
For additional commands, e-mail: users-help@glassfish.dev.java.net

Drinkwater, GJ \(Glen\)

Hi

Glassfish UR1 upon startup throws this error repeatedly for all my EJBs,
taking up a lot of log space, after a while it gives up and says server
startup complete, the application then works fine but I am still need to
know what happening.

[#|2007-02-14T17:19:26.294+0000|WARNING|sun-appserver-pe9.0|
javax.enterprise.system.tools.admin|
_ThreadID=11;_ThreadName=Thread-13;_RequestID=92b51297-a46e-49c7-8d9a-97
147700a245;|
DeferredRegistrationThread.retryItem: deferred registration RETRY failed
after 42812 milliseconds for
"com.sun.appserv:standalone-ejb-module=DP5_EJB_Module,name=bean-pool,typ
e=bean-pool,category=monitor,ejb=DownloadMessageBean,server=server"
(DEFERRING AGAIN)|#]

Any one know whats wrong??

glen

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@glassfish.dev.java.net
For additional commands, e-mail: users-help@glassfish.dev.java.net