chore(ci): measure remote cache phase timings - #1275
Conversation
|
Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Repository YAML (base), Central YAML (inherited), Organization UI (inherited) Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (1)
🚧 Files skipped from review as they are similar to previous changes (1)
📝 WalkthroughWalkthroughThe PR changes the Rust action-cache setting and updates CI qualification. CI checks cache activity, repeats qualification after cleaning on non-pull-request runs, and adds a hosted ChangesRust action-cache qualification
Estimated code review effort: 2 (Simple) | ~10 minutes Merge Risk: ⚪ Minimal · up to This PR changes CI-only cache measurement and gating behavior without changing application or production runtime behavior; no actionable merge-blocking risk remains after normal checks and review. Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Greptile SummaryThe PR updates CI cache qualification to consume mise’s structured statistics reports and adds an archive-cache control for comparative timing.
Confidence Score: 5/5The PR appears safe to merge. No blocking failure remains. Important Files Changed
Reviews (8): Last reviewed commit: "ci: isolate GitHub cache control keys" | Re-trigger Greptile |
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
|
Live normal-hit result on commit
Comparator from the same workflow run using Namespace caching:
The normal mise cache path is correct, but this is not acceptable rollout performance. This PR should remain diagnostic and should not replace Namespace caching until transfer/materialization is optimized. The local cache shape is 3,063 files totaling about 963 MiB; 2,040 files are smaller than 4 KiB, which makes per-blob request overhead a leading suspect. AI-assisted — Tool: Codex; model: unavailable/unavailable; version: unavailable. |
Instruction counts
No instruction-count regression above 1%. Only instruction counts gate. Wall clock is shown for context — on identical hardware it moves 4-20% run to run. Measured by tak — instruction-counted CLI benchmarks, stored in this repository's git notes.
|
Dismissed because a newer commit was pushed; Greptile will re-review the current head.
Dismissed because a newer commit was pushed; Greptile will re-review the current head.
|
Additional cache-shape comparisons:
This changes the rollout bar: the optimized remote cache should be compared primarily with the roughly one-minute GitHub-hosted archive-cache setup, while Namespace remains the attached-volume upper bound. It also validates a tiered deployment model: mount AI-assisted — Tool: Codex; model: unavailable/unavailable; version: unavailable. |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In @.github/workflows/ci.yml:
- Around line 188-196: Update the workflow steps using jdx/mise-action and
actions/cache: set the mise action’s cache input to false, add id cargo-cache to
the explicit actions/cache step, and include github.event.pull_request.number ||
github.ref_name in that step’s key to isolate caches by pull request or branch.
🪄 Autofix
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository YAML (base), Central YAML (inherited), Organization UI (inherited)
Review profile: CHILL
Plan: Pro Plus
Run ID: 96e77dc5-0504-4ff7-bc9d-85ac8ab46ac2
📒 Files selected for processing (1)
.github/workflows/ci.yml
Dismissed because a newer commit was pushed; Greptile will re-review the current head.
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
mise 2026.8.5 production qualification
Compared with mise 2026.8.4, Cargo improved from 7m37s to 2m40s (about 65%), while the complete job improved from 8m17s to 2m51s (about 66%). The GitHub Actions-cache comparator in this run had a cold miss because the updated base changed the The remote cache is substantially improved, but it is not yet competitive with the warm archive cache. This PR should remain open. AI-assisted — Tool: Codex; model: unavailable/unavailable; version: unavailable. |
8cb0fcc to
f564323
Compare
Dismissed because a newer commit was pushed; Greptile will re-review the current head.
mise 2026.8.6 production qualification
Compared with mise 2026.8.5, Cargo improved from 2m40s to 1m48s (about 33%), while the complete job improved from 2m51s to 1m57s (about 32%). The same-commit GitHub Actions-cache control was rerun after seeding its new The reflink materialization release is a substantial improvement, but the remote action cache remains about 62 seconds slower at Cargo and 38 seconds slower end to end than the warm archive control. This PR should remain diagnostic rather than replacing the current Namespace cache. The next material improvement should target packed or bundled remote artifact transfer instead of additional per-file materialization tuning. AI-assisted — Tool: Codex; model: unavailable/unavailable; version: unavailable. |
f564323 to
b8a382b
Compare
Dismissed because a newer commit was pushed; Greptile will re-review the current head.
jdx
changed the title
mise 2026.8.8 phase qualification
The report isolates the bottleneck: prefetch occupies almost the entire cache session, and 3,076 independent blob requests accumulate 6.6 hours of concurrent transfer time. In contrast, the report attributes only 4.27s to local CAS writes and 2.12s to materialization. The next cache performance PR should implement packed/bundled blob retrieval; more local materialization tuning is unlikely to move the result materially. The Namespace production build completed in about 44s on this run; the GitHub Actions cache control completed its build step in about 2m28s. The remote action-cache Cargo build took 8m46s on the current base, so it remains diagnostic rather than ready to replace the production cache. AI-assisted — Tool: Codex; model: unavailable/unavailable; version: unavailable. |
Dismissed because a newer commit was pushed; Greptile will re-review the current head.
0634c55 to
44e9653
Compare
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 1 potential issue.
❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.
Reviewed by Cursor Bugbot for commit 44e9653. Configure here.
Dismissed because a newer commit was pushed; Greptile will re-review the current head.
1 participant
Summary
Why
The previous qualification established correctness and basic restore behavior, but it could not explain where the remaining time went. The structured report separates session, prefetch, manifest lookup, action lookup, blob transfer, local CAS write, and materialization time while recording request counts and restored volume.
That data will show whether the next optimization belongs in the client, server, transfer protocol, or local materialization path. In particular, the per-blob request count provides the baseline for evaluating packed blob transfers.
Validation
mainhk run check --no-stage .github/workflows/ci.ymlactionlint .github/workflows/ci.ymlgit diff --checkAI-assisted — Tool: Codex; model: unavailable/unavailable; version: unavailable.
Note
Low Risk
Changes are limited to CI workflow configuration and observability; they do not affect application code or production runtime behavior.
Overview
CI now measures mise’s production Rust action cache using versioned JSON stats (
MISE_TASK_CACHE_STATS_REPORT) instead of scraping log lines, bumps the qualification job to mise 2026.8.9, and setsrust_cache = trueon the qualify task (dropping{ verify = true }).The cache-qualification step validates report shape with
jq, requires cache hits, and on pull requests still requires remote prefetch and non-zero downloaded bytes; non-PR runs keep a post-cargo cleanlocal pass for a second report. It writes hit/miss and phase-timing tables to the job summary and uploads raw JSON as 30-day artifacts. A newgithub-cachejob onubuntu-latestruns a standardactions/cacheRust build and records exact hit vs build time for comparison, andfinalnow depends on that job.Reviewed by Cursor Bugbot for commit b9f15c7. Bugbot is set up for automated code reviews on this repo. Configure here.