Upstream mode that combines responses

[xylagbx]

Prerequisites

• I have checked the Wiki and Discussions and found no answer

• I have searched other issues and found no duplicates

• I want to request a feature or enhancement and not ask a question

Description

What problem are you trying to solve?

Finding the fastest IP address can be time-consuming, and due to network fluctuations, the IP address obtained through testing may not be the true fastest one.

I use a proxy tool called Surge that initiates TCP handshakes to all IP addresses for domains with multiple IPs simultaneously and selects the fastest connection for subsequent requests. This logic ensures that when accessing websites with a large number of balanced nodes, the fastest node in terms of response speed can always be used, and problematic nodes can be bypassed directly. This can save time for AdGuardHome to test the fastest IP address.

Proposed solution

AdGuardHome can query all DNS servers, merge all query results, and immediately return them without testing the fastest IP address.

[ainar-g]

Hello. AdGuard Home only does that when a user selects “Fastest IP address” upstream mode, so I'm not quite sure what you're asking?

[xylagbx]

I'm sorry for not making it clear earlier. My suggestion is for AdGuardHome to introduce a new upstream mode that would query all available DNS servers, combine the results, and give back all IP addresses.

[ainar-g]

I see. In what order would the answer records be in the combined response? And what if two upstreams respond with valid data, but the third responds with NXDOMAIN or no data?

[xylagbx]

When combining responses from multiple upstream servers, there is no need to worry about sorting the responses. If two of the upstream responses contain valid data, but the third response returns an NXDOMAIN or no data, it is fine to simply return the valid data.

The following is a description of Surge's TCP concurrent connection feature:

For domains with multiple IP addresses, Surge will directly initiate TCP handshake with all IP addresses concurrently and select the connection that completes the handshake the fastest for subsequent requests.

There is no need to worry about increasing the overall network load with this behavior, as the cost of TCP handshake is very small, and concurrent testing will only be performed when accessing the domain for the first time. Subsequent requests will use the preferred IP based on the previous test results, and other IPs will only be attempted if there are issues.

This logic ensures that when accessing websites with a large number of balanced nodes, the fastest responding node can always be used, and problematic nodes can be bypassed directly.