EAP6.1 How to choose standalone vs managed domain HA needed

Latest response

We have to install a production environment with the following architecture:
load balancer
--> NODE 1 HTTPD --> NODE 1 EAP 6.2
--> NODE 2 HTTPD --> NODE 2 EAP 6.2

  • Load balancer redirects traffic to node1/2 httpd. Apache nodes must connect indiferently to any of the active Jboss nodes.

and we have the following requirements:

1- High availability.
- If one node goes down, other node serving all requests
- Not lose of service during deployments
2- Independent deployments in each node
3- Memory data replication. HTTP Session available in both nodes
4- External load balancers. Two HTTP servers in front the application server

We were thinking about implementing a managed domain, but some doubts came out to our mind:

1- Is it worth to use a cluster standalone mode in HA?

2- If we go for the managed domain thinking in terms of scalability
- Do we need a cluster as well for accomplishing REQUIREMENT 3&4?
- Do we have to create two groups in order to do independent deployments in both machines?
- What happens if the domain controller goes down? Is it possible to configure the other node to become the domain controller?

Thanks a lot!

Responses

Hello,

In regards to your inquiries:

Is it worth to use a cluster standalone mode in HA?

This is completely up to you. From a functionality perspective, two standalone nodes with function the same as two managed servers within a domain. Domain mode adds the management capabilities not found in standalone mode. Domain mode allows one to manage a whole domain (host controllers, managed servers, etc) with one domain controller.

Do we need a cluster as well for accomplishing REQUIREMENT 3&4?

Clustering will be required for #3. Only clustered instances can replicate data (HTTP web sessions, stateful EJB, etc) throughout the cluster.

Do we have to create two groups in order to do independent deployments in both machines?

Yes. When using domain mode, deployments are pushed out to a whole server group. If you want a deployment on one instance, but not the other, you will need to use separate server groups.

What happens if the domain controller goes down?

When a domain controller goes down, the host controllers/managed servers will continue to operate as usual. A message will be logged stating that the connection to the domain controller has been lost. Note that the DC does not have to be up and running 100% of the time while your instances are up and running. The DC is really only needed when initially starting instances and when pushing out changes to the domain.

Is it possible to configure the other node to become the domain controller?

See [1] for a high availability solution for domain controllers.

-Jonathan

[1]https://access.redhat.com/site/solutions/255963