How do we configure the mod_cluster from multipl apache severs and being loadbalanced by a hardware load balancer in the front and targeted for multiple JBoss instances.
The architecture is going to be like below
H/W Load balancer is going to distribute the requests to say 4 Apache web servers configured in two physical boxes running on a flavour of Redhat linux.
All the four apache servers would have the mod_cluster set up in them and they will serve static content and sends through dynamic content requests to 2 Jboss instances configured in two physical boxes running on EAP 5.1.X.
Now the queries are
1) How to configure this with web-tier clustering if possible ( in case one web server goes down the other should be able to handle the failed over requests)
2) Where to do SSL offloading
3) We would continue with Sticky session as it is required for JBoss clustering. Do we need to configure anything specific if we need sticky session availability.
4) Is there any limitation with the above set up in terms of serving and load balancing requests together with fail over capability.
Appreciate your ideas and inputs together with references if possible
Thanks Jean for your swift reply. I am trying to put some practical questions in Enterprise app set up with mod_cluster. Hope this would help many people on this forum. Sorry for amny questions. Appreciate your support. I would post the results on this forum once I test.
With the above answers,Does that mean,
1) even if one of the webservers that are running are down (because it is crashed or for some other reason) , what would be the impact?
2) Do all those requests be served by another server automatically or should that be configured somewhere?
3) Do we have any difference of configurations for setting with mod_cluster withClustered JBoss nodes and non clustered JBoss nodes?
4) How do we redirect a https requests through mod_cluster to non-clustered nodes.
Say I have 4 servers where 4 different instances of JBoss running on 4 physical servers. Two of them serving as Clustered and say there are 4 to 5 apps are deployed in that. The other two instances are running with two other application one on each and neither the apps not the JBoss instances are clustered. However, our interest is to use same mod_cluster in all the web servers and get all the requests through mod_cluster. The request could be to Clustered servers( which we knoew the config) or it could be for non-clustered instances.
To put in simple can I have redirection mechanism through mod_cluster for non clustered nodes.
5) Do we have any recommendations for ype of MPM in APache when going ahead with mod_cluster. Is it prefork or worker? If so is there any thing specific to MPM to be configured in mod_cluster.
6) Where can I enable the log to clearly see and understand and validate if mod_clustering is distributing the requests properly or not.
Thanks for your time support, Jean!
1) alll httpd instances are handled the same way by mod_cluster node logic, the impact is the impact your hardware balancer has.
2) see 1/
3) Failover with clustered jboss nodes you can define load balancing groups that correspond to buddy replicated nodes.
4) You can use AJP or ProxyPreserveHost or define the
proxyPort in the connector.
5) borh worker and prefork are OK for high perfs worker is recommanded.
6) mod_cluster uses httpd log logic: directive LogLevel