[GuillaumeLagrange]

Our Actions cache is capped at 10 GB, but a CI run that misses the cache saves over 16 GB. Each run evicts the previous run's caches, so the next one misses again. That's why memtrack-benchmarks on #531 found no cache 15 minutes after saving it.

Most of those 16 GB are copies of the same dependencies:

• The 40 bpf-tests jobs each saved their own ~440 MB cache. Shards on the same OS build the same thing, so now one shard per OS saves it: 5 caches instead of 40.
• The ubuntu-26.04 jobs (lint, tests, exec-harness-tests, basic-run-test, bpf-tests) also build mostly the same dependencies. A new prime-rust-cache job builds them once into a shared cache, and those jobs only read from it.

A run that misses the cache now saves about 6.4 GB instead of 16 GB.

The jobs reading the shared cache wait for prime-rust-cache. On a cache hit it finishes in about a minute. The slowest job, memtrack-benchmarks, doesn't wait for it, so overall CI time shouldn't change.

[codspeed]

Merging this PR will not alter performance

2 warnings

⚠️ Walltime results may be unreliable

Some benchmarks used the Walltime instrument outside CodSpeed Macro Runners, which can make the results inconsistent.

Learn more about Macro Runners

See affected benchmarks

⚠️ 9 benchmarks are dominated by system calls

System calls cannot be simulated consistently and are left out of the measure. The reported value understates the real cost. Switch to the Walltime instrument to measure them.

Learn more about system calls

See affected benchmarks

✅ 41 untouched benchmarks
⏩ 6 skipped benchmarks¹

---

_{Comparing ci/share-bpf-tests-rust-cache (229413f) with main (5c86dc3)}

Footnotes

1. 6 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[greptile-apps]

Confidence Score: 4/5

[Medium impact] The changes affect cache use and job ordering across a substantial part of CI.

The PR appears safe to merge, but nine jobs will miss the intended shared cache until their extra key is removed.

Findings

1.  Readers miss the shared cache ▶

Fix with agent prompt

### Issue 1
.github/workflows/ci.yml:145-146
`basic-run-test` and `bpf-tests` still pass `cache-key: ubuntu-26.04` when selecting the shared cache. `shared-key` replaces the job name, but the extra `key` is still appended. The primer passes no extra key, so these nine jobs look under a different cache prefix and rebuild dependencies despite waiting for the primer.

Clear `cache-key` for both shared-cache readers, or have `install-rust` omit it whenever `shared-key` is set.

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

Summary

This PR reduces duplicate Rust cache saves:

• Ubuntu 26.04 CI jobs reuse one prepared Rust cache.
• BPF test shards stop saving a separate cache for every test.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  P[Prepare Ubuntu 26.04 cache] --> C[Save shared cache]
  C --> R[Lint, tests, exec-harness tests]
  P --> B[Basic-run and eight BPF shards]
  B --> K[Extra key selects a different cache prefix]
  K --> D[Rebuild dependencies]

Loading

_{Reviews (1) · Last reviewed commit: "ci: prime one shared rust-cache for the ..." · Reviewed by Greptile}

[not-matthias]

nice work! 🔥 🚀