Buildx gRPC ResourceExhausted error – looking for guidance / best practices

[RushiChandalia]

Hi folks 👋

We’re running into a recurring issue with docker/build-push-action@v5 + Buildx/BuildKit and wanted to start a discussion to see if others have hit the same problem and whether there are recommended solutions or best practices.

❌ The issue

Our builds occasionally fail with the following error:

ERROR: failed to build: failed to receive status:
rpc error: code = ResourceExhausted
desc = trying to send message larger than max (26355782 vs. 16777216)

From what we understand, this is due to BuildKit’s internal gRPC message size limit (~16MB) being exceeded.

🧱 Context

• Using docker/build-push-action@v5
• GitHub Actions runners
• Pushing images to internal AWS ECR
• Secrets are passed via --secret (JFrog token, AWS creds, etc.)
• No custom buildx builder config
• Default BuildKit behavior

Example snippet (simplified):

• uses: docker/build-push-action@v5
  with:
  context: ./service
  file: ./service-package.Dockerfile
  tags: my-image:latest

Using ChatGPT and other resources I was able to get this recommended fixes:

1. Increase gRPC limit - Set environment variable in the buildx action:

env:
  BUILDKIT_STEP_LOG_MAX_SIZE: 50000000

2. Disable provenance/sbom - Reduces metadata size:

with:
  provenance: false
  sbom: false

For Solution 1 I am not able to find any documentation on how to use that env variable and where to use it.

And for solution 2, this feels more like a workaround than a real solution. (Because we want to keep metadata with the build)

🤔 What we’re hoping to understand

Is there any supported way (now or planned) to:

chunk or stream large metadata
raise the gRPC limit safely
avoid sending large attestations over gRPC entirely?

Are there recommended best practices for:
large repos / monorepos
builds with many dependencies

Is disabling SBOM/provenance the intended solution in these cases, or just a known limitation today?

Appreciate any insights from maintainers or others who’ve run into this 🙏
Happy to provide more details if useful.

Thanks!