1. The gateway that is connected to is whatever URL that the users are accessing. So if this URL is a user friendly DNS entry, then whatever the underlying IP is for this DNS is which gateway is pointed.
2. No. There is not automatic failover of gateways. You have the choice of manually repointing the URL to point at one of the workers, or the better option is to deploy a load balancer that handles the incoming traffic and knows how to route the traffic to any of the available gateways.
3. The repository typically will have automatic failover. So if worker1 fails, then worker2 repository will become active.
See H.A. for more info, Distributed and High Availability Tableau Server Installations
Thanks for the responses.
1. Regarding gateway- How the request is serviced? Suppose I have all components configured in 3 workers and Tableau server receives the request like login , opening the report and do a manual refresh. In this case, which gateway will take the request and components related to which worker will be used?
2. Do we have any open source load balancers or anything which is best supported by Tableau?
3. I want to do a manual switching of repository? Do we have any commands to fulfill this requirement?
Also, is there any possibility of mapping group IP's (Primary + workers) to point one DNS entry so that even if gateway fails, the other gateways can take request.
Can you please provide me the info on above asked points. Thanks!!
1. The initial request is serviced by the gateway that the user friendly DNS points at. Any internals between Tableau processes (i.e. vizql, dataserver, data engine) is handled by Tableau's internal apache load balancer and any of the gateways will be leveraged.
2. I don't know of any open source load balancers. We use nginx and netscaler. Both work
3. There is a tabadmin command to do a manual failover of the repository. tabadmin commands
Extra. No, at least not to my knowledge. What I understand from our team that works with DNS, is that each DNS is mapped to a singular IP.
Thanks Jeff for clearing my queries. They were really helpful.