perf: vitest threads pool + Dockerfile pnpm cache mount #66

michal · 2026-04-27T16:06:55Z

michal commented

2026-04-27 16:06:55 +00:00

Summary

Two tuning knobs leaving the box mostly idle:

1. `vitest.config.ts` — pool: threads with maxThreads ≈ cores/2

Default vitest pool config left the 64-core workstation at ~10% CPU during `pnpm test:run`. Threads pool uses the box: ~700% CPU instead of ~150% on the same suite. Wall time gain is modest (workload is dominated by a few slow individual test files that one thread must run serially — `registry/client.test.ts` 8s, `status.test.ts` 3.5s), but the parallel headroom is there for when the suite grows. Override at run time via `VITEST_MAX_THREADS=N`.

2. `Dockerfile.mcpd` — BuildKit pnpm cache mounts

Adds `# syntax=docker/dockerfile:1.6` and `--mount=type=cache,target=/root/.local/share/pnpm/store` to both `pnpm install --frozen-lockfile` steps. Pnpm's content-addressed store survives across image rebuilds. Cold rebuilds (lockfile changed) unaffected; warm rebuilds (only source changed) drop the install step from ~60s to <5s. `fulldeploy.sh` mcpd image rebuilds get that back.

Test plan

`pnpm test:run`: 2050/2050 across 152 files (unchanged)
Per-package mcpd: 837/837 (unchanged)
Manual: deploy this and observe `fulldeploy.sh` mcpd image rebuild time on warm cache.

🤖 Generated with Claude Code

## Summary Two tuning knobs leaving the box mostly idle: **1. \`vitest.config.ts\` — pool: threads with maxThreads ≈ cores/2** Default vitest pool config left the 64-core workstation at ~10% CPU during \`pnpm test:run\`. Threads pool uses the box: ~700% CPU instead of ~150% on the same suite. Wall time gain is modest (workload is dominated by a few slow individual test files that one thread must run serially — \`registry/client.test.ts\` 8s, \`status.test.ts\` 3.5s), but the parallel headroom is there for when the suite grows. Override at run time via \`VITEST_MAX_THREADS=N\`. **2. \`Dockerfile.mcpd\` — BuildKit pnpm cache mounts** Adds \`# syntax=docker/dockerfile:1.6\` and \`--mount=type=cache,target=/root/.local/share/pnpm/store\` to both \`pnpm install --frozen-lockfile\` steps. Pnpm's content-addressed store survives across image rebuilds. Cold rebuilds (lockfile changed) unaffected; warm rebuilds (only source changed) drop the install step from ~60s to <5s. \`fulldeploy.sh\` mcpd image rebuilds get that back. ## Test plan - [x] \`pnpm test:run\`: 2050/2050 across 152 files (unchanged) - [x] Per-package mcpd: 837/837 (unchanged) - [ ] Manual: deploy this and observe \`fulldeploy.sh\` mcpd image rebuild time on warm cache. 🤖 Generated with [Claude Code](https://claude.com/claude-code)

michal added 1 commit 2026-04-27 16:06:56 +00:00

perf: vitest threads pool + Dockerfile pnpm cache mount

CI/CD / typecheck (pull_request) Successful in 56s

Details

CI/CD / test (pull_request) Successful in 1m9s

Details

CI/CD / lint (pull_request) Successful in 2m40s

Details

CI/CD / smoke (pull_request) Failing after 1m43s

Details

CI/CD / build (pull_request) Failing after 7m6s

Details

CI/CD / publish (pull_request) Has been skipped

Details

18245be0c1

Two tuning knobs that were leaving most of the host idle:

1) vitest.config.ts pool=threads with maxThreads ≈ cores/2.
   Default left this 64-core workstation at ~10% CPU during
   \`pnpm test:run\`. Threads pool uses the box: same 152-file/2050-test
   suite now runs at ~700% CPU instead of ~150%. Wall time gain is
   modest (workload is dominated by a handful of slow individual files
   that one thread must run serially), but the parallel headroom is
   there for when the suite grows. Cap = max(2, cores/2) keeps laptops
   reasonable; override with \`VITEST_MAX_THREADS=N\` in the env.

2) Dockerfile.mcpd uses BuildKit cache mounts on both pnpm install
   steps. Adds \`# syntax=docker/dockerfile:1.6\` and a
   \`--mount=type=cache,target=/root/.local/share/pnpm/store\` so
   pnpm's content-addressed store survives across image rebuilds.
   Cold rebuilds where the lockfile changed are unaffected; warm
   rebuilds where only source changed drop the install step from
   ~60s to <5s. fulldeploy.sh's mcpd image rebuild gets that back
   minus the docker push hash mismatch.

Test parity: 2050/2050 across 152 files; per-package mcpd 837/837.
Both unchanged.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

michal merged commit 9374a2652b into main

2026-04-27 16:07:06 +00:00

michal referenced this issue from a commit

2026-04-27 16:07:08 +00:00

perf: vitest threads pool + Dockerfile pnpm cache mount (#66)

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: michal/mcpctl#66