Mesh settings clarification and bugs

[TheOneOgre]

Hello all. I have been using CodeProjectAI for a couple years now for processing images from my BlueIris server. This has worked great and I love using it.

The only thing that seems to be severely lacking in either development, discussion, or knowledgebase entries, is Mesh support.

I have 2 servers running in mesh with each other, and while they do occasionally send requests to each other, the mesh json configuration seems to be all over the place.

For one, I tried following exactly what the documentation says:

An example would be a Raspberry Pi and a Macbook Pro. The Pi might have AcceptForwardedRequests false and AllowRequestForwarding true, and the mac could have AcceptForwardedRequests true and AllowRequestForwarding as false. This provides the Pi with help if needed, without worrying about the mac sending unwanted load.

So on my primary server with the GPU (which I prefer requests go to) Has AcceptForwardedRequests true and AllowRequestForwarding false. This actually breaks mesh as, in the Mesh tab, it shows that the server will not accept or forward requests. Changing AllowRequestForwarding to true, reenables the acceptance, but also reenables forwarding.

On the second server, AcceptForwardedRequests is false and AllowRequestForwarding true, but this shows as both it accepting and sending requests.

This is a bug as, it should work exactly like the documentation states, but it does not.

Secondly, I wish there was an option like PreferMeshServer so that my secondary server will prefer sending requests to the primary server, and only fallback to itself processing if the primary one is unreachable/down for some reason.

Any way for this to be implemented, and as for the bug mentioned, is this a misconfiguration or misunderstanding on my part, or can this be fixed?

Thanks to all who work on this project!

[TheOneOgre]

Just as an update to this, changing "HeartbeatInactiveTimeout": from "00:00:15" to "00:02:15", keeps the timing open long enough for my primary server to stay as preferred. Since my application (BlueIris) was only sending requests about every 2 minutes at worst, and 20 seconds at the soonest (static object detection), the timings would drop back to 0ms and the server would be marked as "Active: false".

This change now makes sure that the faster server (my main server with a GPU) is preferred and the connection stays open.