How to set up a local RBE for 3 Windows machines using Siso?

[曾衣其]

Hi all,

I have three Windows 11 machines on a local network and I want to use them to speed up my Chromium builds.

I’m planning to set up a local RBE (Remote Build Execution) cluster so these 3 machines can build Chromium together using Siso.

I have a few practical questions:

1. RBE Backend: For a small local Windows cluster, which RBE implementation (Buildgrid, Buildbarn, NativeLink, etc.) works best with Siso on Windows?

2. Siso Config: What are the specific gn args or environment variables needed to point Siso to my local gRPC RBE endpoint instead of the default Google infrastructure?

3. Environment Sync: Do I need to manually ensure that all three machines have identical Visual Studio/SDK paths and toolchains installed, or does the RBE/Siso setup handle this isolation?

If anyone has a working config for a small local Windows cluster, I'd appreciate any pointers.

Thanks!

[Sidhin Thomas]

In case you haven't seen this before - https://chromium.googlesource.com/chromium/src/+/main/docs/linux/build_instructions.md#non-google-rbe

This should work for windows also. 

Frankly I didn't have a lot of success with it. I didn't pursue this since my system builds chromium in 1.5 hours and that's okay for a full build every couple of days for me. 

Please let me know if you see very significant gains with 3 machine clusters using RBE. 

--
--
Chromium Developers mailing list: chromi...@chromium.org
View archives, change email options, or unsubscribe:
http://groups.google.com/a/chromium.org/group/chromium-dev
---
You received this message because you are subscribed to the Google Groups "Chromium-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to chromium-dev...@chromium.org.
To view this discussion visit https://groups.google.com/a/chromium.org/d/msgid/chromium-dev/6f5fc373-147a-4023-9301-698c8708a572n%40chromium.org.

[Junji Watanabe]

In addition to RBE workers, you need to prepare RBE API servers to schedule requests to workers and CAS servers that host remote caches. 

I don't know the power of the three machines. However, the performance gain is visible only when you run many remote compiles in parallel (e.g. 1k+ remote compiles) because each compile becomes "slower" due to the overheads.

To view this discussion visit https://groups.google.com/a/chromium.org/d/msgid/chromium-dev/CAGPuNTWvf%2BZ6k4F2eTKbE5O_9G%2B0%2B5-YpSLn_nXuhT%3DdZ5voSg%40mail.gmail.com.

[Sidhin Thomas]

Inspired by the thread. I tried another hand on RBE and got native link working. The benefits from caching are huge vs ccache. But in order to get it working I had to hard code the container-image from google.star while building the worker image. I'll update it every now and then.

Any idea, how much and what kind of issues can I expect due to toolchain drift when using the RBE for dev workflows? C++ in general is ABI compatible so I am assuming it should be safe. 
(I am basically ignoring toolchain as a match condition for workers)