Does Codecov efficiently stream coverage data for large binary monorepos?

Answer to the question

Codecov does not currently stream or chunk coverage reports for large binary monorepos. The Codecov CLI bundles the entire coverage report into a single HTTP request, which can cause CI timeouts or memory pressure when the report exceeds the limits of the runner or the CI provider’s upload size threshold.

Confirmed behavior

• The Codecov CLI (or API) zips the coverage report and sends it as one multipart/form-data payload.
• There is no built‑in streaming or chunked transfer encoding for coverage uploads.
• Upload failures or truncation will leave the coverage bottleneck view incomplete or inaccurate.

Likely explanation

Large binary assets inflate the coverage report size because the instrumentation framework (e.g., lcov, nyc, coveralls) records every file touched by the tests, even if it is a binary. When the report grows beyond a few megabytes, the single HTTP request can exceed CI runner memory limits or the CI provider’s maximum request size (GitHub Actions: 250 MiB; GitLab CI: 100 MiB; self‑hosted runners: depends on configuration). The CLI’s --timeout flag only retries the request; it does not change the payload size.

Because the bottleneck view is generated after the report is fully processed, any upload issue directly affects the accuracy of low‑coverage file identification and regression detection.

Practical steps for your monorepo

1. Measure report size. In your CI job, add a step:

   echo "Coverage report size: $(du -sh coverage/ | cut -f1)"
   

2. Filter binary files. Configure your coverage tool to exclude binaries. For nyc:

   nyc --exclude "**/*.bin" --exclude "**/*.png" npm test
   

3. Compress before upload. The CLI already zips the report, but you can reduce size further by running:

   gzip -9 coverage/coverage-final.json
   

4. Use a larger CI runner. If you are on GitHub Actions, switch to a runner: self-hosted with more memory or use a larger hosted VM (e.g., ubuntu-latest with 8 GB RAM).
5. Retry logic. Add a simple retry loop around the Codecov upload:

   for i in 1 2 3; do
     codecov || sleep 10
     [ $i -eq 3 ] && break
   done
   

6. Verify the bottleneck view. After a successful upload, check that all expected files appear in the Coverage bottleneck view. If files are missing, the report was truncated.

Missing diagnostic detail

To give a more precise recommendation, please share the approximate size of your coverage report and the CI provider you are using (e.g., GitHub Actions, GitLab CI, CircleCI, self‑hosted runners). This will help determine whether the current runner limits are the root cause or if additional filtering is needed.