We are currently using Crowd to authenticate Jira and Confluence.
Recent development on our core applications includes the use of Crowd for authentication. This put me in a situation where I would like Crowd to be (more) redundant.
Current set up is on 1 virtual server. The hardware this server is on is currently unknown and I have a request put in to our ASP in regards to information on the hardware (e.g. at least have the most error-prone components redundant).
Having suitable hardware is just step 1. I would really like to have the Crowd installation redundant. I've done some research and found the following links:
It seems to me that there are some issues in regards to the setup - one example is that other Crowd installations using the same db must be turned off when adding a new application to Crowd. This is not the biggest of issues as we do not add that many applications during a year - but still...
Solution is either to use hardware or software in front of the two (LB) and have the related applications use the cluster ip-address instead of the specific server ip address we use today. My question is how to avoid the hardware/software solution in front (the actual LB) becoming the single point of failure? Yes, we can have two of this as well, but there is something called costs 
Can anyone tell me:
- is it possible to have a single LB instance and have traffic routed directly to the ip address of one of the servers behind the LB if it goes down?
- Will Jira/Confluence/other applications manage this switch?
Or is my only solution to duplicate the LB instances?
