Please provide a way to see "game contributions" statistics per PC, not only per user

[Technologov]

Hello,

BOINC (WorldCommunityGrid) allows me to see stats per PC and per user.
This is very important for data-center guys with 100+ PCs, like me.
Now, despite having 100 PCs, only 1 contributes, because only 1 has an RTX GPU.
But in the future I will have a lot of RTX GPUs. Contributing CPUs to Leela is a waste of watts / power.

I would like to see similar kind of statistics for Leela training servers, where I contribute games.
I would like to break down which PCs were active, which were not, and how much each PC contributed games so far. (this could let me detect GPU issues, cooling issues, network issues and other issues easily)
This feature request is mainly about "website", but also about "client" application, where I want to see statistics locally.

Even better idea: Make a Boinc-Leela, so I could run Leela client as a BOINC application.

1. This will simplify management, because BOINC has fairly advanced management tools built-in and developed over a decade, for both GUI and CLI, both Windows, Mac and Linux (and Android).
2. This will gain a lot of free marketing. = More Leela contributors.

Here:
https://lczero.org/
and here:
https://lczero.org/active_users

-Technologov, 7.2.2019

[Technologov]

Developers !!! Please provide much better website statistics. I wanna know how many of my PCs are connect to Leela server, and which versions (client, engine, run) they are running. Else it will be hard for me to debug a problem if one of my PCs stays on old version, or has a hardware failure.

I need the following data, per PC:

• client version
• engine version
• run (=1 or =2)
• GPU + VRAM
• hostname (much easier to debug, if things go wrong)
• IP address
• last connect time
• amount of # GPUs
• OS + version (Windows, Linux, etc...)

[mooskagh]

If you need to track clients separately, the easiest is to have a separate username per client.

Currently there's no way to get GPU and VRAM out of engine (and it would be hacky to add that), and we don't store ip address, hostname or OS.
Exposing many of those values (especially ip address and hostname) may also cause privacy considerations.
Some of the listed items are also not straightforward to implement for all configurations (CUDA and BLAS, Linux and Windows etc).

What is less complicated to do is client/engine version, run id and last connected time, as that's what server already has.

[Technologov]

If you need to track clients separately, the easiest is to have a separate username per client.

It's not easiest at all, if I need to create 10 separate users and manage them.

[mooskagh]

Maybe it's not easy (but for me it does look easy), but it's surely easier than to modify engine-client-server chain, so it does look "the easiest" for me.

There's no need to create user, it's created automatically on first usage.
Also you can modify the script that starts training to include e.g. IP address in the username, and they all can have the same password. That way you don't have to manually modify script at every instance that you run.

[AlexisOlson]

If I can get a connection string to query the lczero server, I could write a report similar to the !contributions report (which scrapes the lczero.org/networks/ page) that could include existing data items like version, engine, run, last uploaded game etc. for each user.

The other data items would still have to be handled using differently named users unless someone is willing to do all the coding to support the features Technologov is requesting.

I know I'd love to be able to see my contributions over time and leaderboards that aren't stale most of the time.

[madformuse]

@Technologov I'm working on an integration with BOINC using the boinc-server-docker images. It would be good to catch up some time to talk about how you would see it working.

[Technologov]

Sure. Let's talk. My telegram and WhatsApp = +972-58-4433150 My email is AL4321@GMAIL.COM

…

On Mon, Apr 22, 2019, 13:04 madformuse ***@***.***> wrote: @Technologov <https://github.com/Technologov> I'm working on an integration with BOINC using the boinc-server-docker images. It would be good to catch up some time to talk about how you would see it working. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#35 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AALJPNDPDFBY5KGV54RIPTDPRWLVDANCNFSM4GU5TFFQ> .