I've been playing with Oozie for a while. Since it looks like an important part of the picture (important in the sense that it would be good that it's available as much as it can be), I'm considering some "failover" and/or HA configurations. Since I'm not able to find more detailed information on some specifics, I'd like to ask a couple of questions:
a) In the case when there's only one Oozie role, is it possible to move it to another server if the first one fails? If I understand correctly, Oozie can use a shared database to store it's data, so the manual migration looks ok to me. Is this procedure described somewhere (if it's possible to do it)?
b) In the case of HA implementation, as described in the docs (https://www.cloudera.com/content/www/en-us/documentation/enterprise/latest/topics/cdh_hag_oozie_ha.h...), is the Load Balancer mandatory? It should be possible to make clients intelligent and aware of primary/secondary Oozie server to remove the LB out of the architecture, but what about setting the LB address (it's mandatory to set one during the HA configuration)? Is it ok to put the address of one of the Oozie servers (I'm aware that then this isn't a real HA configuration per say, but ..)?
Another question related to this ..
While testing several different scenarios, I've tried to simulate a failure of oozie service without HA/second instance. The idea was to simply reasing the oozie role to another server if the primary fails. Steps for doing this are simple: add oozie role to another server, stop and delete the first (failed) instance, start the new one and restart Hue to pick up the config. This seems to work fine - all jobs/workflows are automatically continued, so I believe this can serve as a manual failover scenario in case of oozie server failure. Am I missing some relevant info or is this a valid assumption?