For your question # 1 fail-over. You do not have gateway on workers. If gateway fails, your cluster is down so no HA. You do have data engine on primary and worker1. If Data engine fails, queries running via proxy through the Data engine process will fail, resulting in a failed view rendering. However subsequent requests, including a retry of the failed operation, should succeed as long as a working Data Engine still exist and can accept rerouted request. So you do have fail-over for data engine that handles request for data source connections.
I think that your configuration is more than enough for the numbers of users. Both backgrounder and Vizql server are CPU intensive. You have total 16 cores that should be good for 2,000-3,000 users. NetApp's 16 core handles 3,500 users today with about 50-60% CPU average. You can monitor the server resources and may want to add more memory if memory goes above 70% average. You can monitor CPU & memory either from standard windows tool or use TabMon.
Server 9.2.4 would give you a lot of more HA features. Details @ http://www.tableau.com/sites/default/files/media/whitepaper_9.0hamissioncriticalrapidfire_eng_1.0.pdf
Thanks Marc.If i have gateway process on all worker nodes along with primary node, is it possible that if the gateway process on Primary fails, can i send the url of 1 of the worker node to users so that they can connect to Tableau server?
Thanks for all your support.
1 of 1 people found this helpful
I am glad that you do configure gateway process on all nodes that is recommended configuration for 8.1 and late versions. So what happens when a Gateway process fails? If other Gateway processes remain running, requests made to those working Gateways will be processed normally. However, any requests received by the failed Gateway will continue to fail despite the presence of other functioning Gateways. Failed Gateway processes automatically restart; so as long as the computer itself is working, the Gateway process will relaunch and resume serving requests. It is not necessary to send worker’s url to users as auto-restarting of Gateway process should not take long. Letting users to access Tableau via worker url may cause more confusions (SSO may not work either)….
If your risk tolerance to individual Gateway failures is very low, you should consider an external load balancer in front of all gateways.