Tableau Software : Upstream Connect Error or Disconnect/Reset Before Headers

This error indicates a communication breakdown between the Tableau Gateway (which uses Envoy) and a backend Tableau process (like VizQL, Data Server, or Backgrounder).

Essentially, the Gateway tried to pass a request to a service, but the connection was cut before it could receive a response.

This is most commonly caused by a timeout (the request took too long) or a crashed service (the backend process died).

Here is the step-by-step troubleshooting guide for Tableau Server administrators.

Step-by-step Fix

1. Check Server Status (Immediate First Step)

Before changing configurations, verify if a specific service is down.

• Open a command prompt (as Administrator) or terminal on the Tableau Server node.

• Run: tsm status -v

• What to look for: Look for any process listed as “Error” or “Degraded”, particularly:

  • Ask Data (Elastic Server) – A frequent culprit for this specific error.

  • VizQL Server

  • Tableau Prep Conductor

• Fix: If a service is down, try restarting the server: tsm restart

2. Increase the Gateway Timeout

If your dashboards are complex or your data extracts are large, the backend process might simply be taking longer to respond than the Gateway allows (default is often 30-60 seconds for certain internal hops).

You can increase the global gateway timeout using TSM commands:

1. Set the timeout (e.g., to 180 seconds):

   Bash

   tsm configuration set -k gateway.timeout -v 180
   

2. Apply the changes:

   Bash

   tsm pending-changes apply
   

   (Note: This will likely require a server restart.)

3. Check External Load Balancers (AWS/Azure/F5)

If you have a load balancer (LB) sitting in front of Tableau Server, the error might be coming from the LB, not Tableau itself.

• The Rule: Your External Load Balancer’s idle timeout must be higher than Tableau Server’s internal timeout.

• Example: If Tableau is set to process a request for up to 3 minutes, but your AWS ALB is set to timeout after 60 seconds, the user will see an “Upstream Connect Error” after 60 seconds.

4. Specific Known Causes

• “Ask Data” / Elastic Server Failure: If you do not use the “Ask Data” feature, but the Elastic Server process is failing (common in versions 2021.x – 2023.x), it can throw this error on the login or home screen. You can disable it to stabilize the server:

  Bash

  tsm configuration set -k ask_data.enabled -v false
  tsm pending-changes apply
  

• Tableau Cloud Users: If you see this error on Tableau Cloud (hosted by Salesforce), you cannot fix it yourself. This indicates an outage or instability in the Salesforce Hyperforce infrastructure. Check the Salesforce Trust page and open a support ticket.

Summary Checklist for Admins