What's the correct way to configure cluster with load balancing and data synchronization jboss/jbpm/helix/zookeeper

Posted on

- RHEL Server 7.7
- Jbpm 7.26 full installer
- Oracle (I think is working find with de ds configured)

I installed jbpm 7.26 in two hosts trying to configure as a cluster using zookeeper and helix, following the documentation.
But I'm really confused at this time because I can't get the way to make it work.
What I did: compiled with ant install.demo, start it and then stop it. So I can get the package created for wildfly.

First attemp: configure both nodes as helix controllers, and run them with the same command (./bin/domain.sh) using host.xml "properly" configured. This way I can't notice any cluster behaviour, as I can access every node by its own address. And, if one of them got down, the other one dosen't notice.

Second attemp: configure one as a master and the other one as slave. Helix controller is in the master. The last one registers to the master and in the master, I can see the registration (logs). Whit this, I can only access through the master address, but, if it got down, I lost connection, the slave doesn't become master.

What I espect (I'm not sure if this is the correct way to see it):
1. I, as a user, can connect to workbench throug (only) an address, and the cluster (Helis?, Zookeeper?) is responsible of redirecting with one node.
2. If master node get down, the other takes control.
3. When the other node takes control, every data is there.
4. I should be able to add more nodes with the less human intervention as posible.

How, exactly, should I configure vfs-repo (for replication?) and what is the (general) espected behaviour of the cluster?