1 of 1 people found this helpful
1. There is no "active standby" mode in Tableau Server. You're trying to express a state of one or more of the machines here, but I don't understand what you mean. Can you be more explicit?
Each gateway is a built-in Tableau "Load Balancer". You can also balance across these gateways with your own hardware load balancer so that if one of "our" gateways goes down, there are others which are still there to handle requests.
2. Nodes don't communicate with each other - individual processes do. The gateway process also helps figure out what is up and what is down.
3. Assuming you have another gateway running and an ELB to route requests to it, Tableau keeps working. You must get your backup primary up and running and/or restore the primary before ~ hours or some of the components may de-activate when they attempt to do a licensing check and find no primary available to service them. This is all documented in help.
4 That's up to you. All the gateway does if forward requests to the appropropriate process based on what the request needs to do - it doesn't really "behave" in any particular way. Perhaps you can expand your question?
5. You load balance nodes which have a gateway running on them
6. Again, "active standby mode" doesn't mean anything in the Tableau world. Your ELB will route requests to one of our gateways based on whatever rules YOU setup. When a request hits one of our gateways, it is distributed in a simple round-robin fashion to the next available process on any machine.
For the 1st point , yes i have mentioned in my question that i am pointing towards three node tableau cluster system where worker 1 is in active mode and worker 2 is in stand by mode As u have mentioned each gateway is a built in Tableau "Load Balancer" , that means suppose i have a three node tableau cluster where my primary server is running on "x" ip and my other two workers (active - standby mode) are working on "y" and "z" ip , when my client "A" and client "B" logs on to primary tableau server via "x" ip then will the request for client A will be served by worker 1 and client B will be served by worker 2 or vice versa or only worker 1 will serve request ? since i am confused with the term load balancing because in a three node tableau cluster environment worker 1 is in active and worker 2 is in standy node , so how load balancing comes in to picture since only my worker 1 is in active mode and worker 2 in standby mode.
Please tell me what exactly gateway inbuilt load balancer does ?
For "Nodes don't communicate with each other - individual processes do. The gateway process also helps figure out what is up and what is down." As you have told here nodes don't communicate with each other but individual process do. so are you referring here vizql , background and other processes ? A vizql process running in worker 1 can communicates with vizql process running in server 2 ? In what cases they will communicate with each other . Suppose i shutdown my worker 1 server then how gateway process running in both primary and worker 2 will know that worker 1 node is down ?
Please tell me based on what all scenario's gateway should be running in only one node or all the three nodes ? As per me as u have earlier mentioned that gateway is a built in "load balancer" so in a three node cluster system it should be running on all the three nodes for load balancing.
Please tell me for External load balancer do i need a extra hardware other than three node tableau cluster. Also i referred the link Add a Load Balancer to Configure Tableau Server to Work with a Load Balancer. Are this steps to be carried out only on primary tableau server machine or worker 1 and worker 2 also ?
I am a bit confused with this , just need a deep understanding on this ..
2 of 2 people found this helpful
Well, that’s what I mean – there is no such thing as “active” and “standby” workers in Tableau Server. All workers are active all the time. A particular service on a worker might be in standby, (like the repository), but all other services are active. So, what it sounds like you’re calling “standby” is actually “active” – just a service or two on the machine are standby while everything else is active – and needs to be balanced
Please post a screenshot of your machine config (The maintenance page) if you need further clarification.
Thanks Russel for this clarification. And also for External load balancer do i need a extra hardware other than three node tableau cluster. Also i referred the link Add a Load Balancer to Configure Tableau Server to Work with a Load Balancer. Are this steps to be carried out only on primary tableau server machine or worker 1 and worker 2 also ?
We have some performance issues with an existing Tableau server implementation. Please find the architecture diagram below.
Currently there are a lot of performance issues being reported. There are about 5000 Users, 25 sites and a lot of Extracts scheduled. There are network latency reported between worker nodes which are being sorted out.
Could you provide your comment on the architecture and suggest a possible better architecture and distribution of processes with our existing configuration. Will increasing any of the processes in the server help? if so what are the options?
2 of 2 people found this helpful
Hi Vinod -
We try not to give prescriptive guidance about how to setup "your" cluster on the Tableau Forums - that is the realm of Professional Services and (to a certain extent) Technical Support.
That said, 2 VizQLs on a single 4 core (Worker 1) is suspect. Fully half your VizQL load will be going to a machine that is too small from a CPU perspective. I'm guessing you know you need 32 cores, but only have 24 right now?
You also doing anything with your Primary - Why not run a Gateway on it? Doing so won't cost anything. It would allow you to remove the Gateway on Worker1.
I suspect you're also going to have problems when the Data Engine on Worker 2 tries to answer questions while your Backgrounders are running. That's 3 heavyweight processes all hitting a single 4-core.
Hope this gives you something to start with.
3. If the primary server goes down what will happen ?
After an interval that is unknowable to you (between 0 seconds and 8 hours, depending on how long ago the licensing check happened), your cluster will come down.
This is described in the section called "Configure a Backup Primary", in the server admin documentation.
I call this "High-ish availability."
Thanks a lot for the reply. One more question. what are the advantages of placing a gateway in the primary server? I could not find any writeups reagrding this online.
[Sorry, didn't realize the question was for Russell until I had already replied.]
The advantage, is that you're using your primary's processing power. Otherwise, it's sitting there doing nothing (other than making sure your payments to Tableau are current ).
If you're using a relatively small computer or VM, you don't want to run a worker there, but gateway doesn't require much CPU.