fix(openontology): migrate in one transaction, and stop timing out on slow disks - #124
Merged
Conversation
… slow disks CI failed on an unrelated PR when "seeds a package and hydrates it back identically" passed vitest's default 5s timeout. The assertions were fine; the suite is I/O bound and the runner was slow. Two changes. migrate() ran every DDL statement through its own client.execute(), so migration 1's ~30 statements each became a separate durable commit and opening a store paid ~30 fsyncs. Batch each migration into one write transaction instead: locally a fresh migration drops from ~8.2ms to ~5.0ms, and the gap widens as fsync gets more expensive. It also closes a real hole -- a crash part-way could previously leave the schema half-applied while schema_migrations recorded the migration as done, because the statements and the bookkeeping insert were not atomic. Then give the package a 30s testTimeout. These suites drive a real file-backed SQLite database, so their wall time is set by the host filesystem, not by our code. The 5s default is tuned for CPU-bound unit tests and leaves no headroom on a contended runner. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
vu1nz Security Review0 finding(s) in PR #? No security issues found. |
ralyodio
marked this pull request as ready for review
August 3, 2026 13:35
This was referenced Aug 3, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Why
CI failed on #123 — a marketing-copy PR that cannot touch this package — when
packages/openontology/src/libsql.test.ts→ "seeds a package and hydrates it back identically" hit vitest's default 5000ms timeout. Re-running the same commit passed, so it is a flake, and it will keep landing on unrelated PRs until the underlying variance is removed.The assertions were never wrong. The suite is I/O bound and the runner was slow. Locally the whole file runs in ~250ms with the slowest case at 79ms, so CI diverged by 60×+.
What
1. Batch each migration into one write transaction.
migrate()ran every DDL statement through its ownclient.execute(). Migration 1 is ~30 statements, so opening a store cost ~30 separate durable commits — and every test in the file opens several stores.Measured on a local SSD, 25 fresh migrations:
~40% off where fsync is cheap; the win grows as durability gets more expensive, which is exactly the CI case.
This also closes a real hole. The DDL and the
schema_migrationsbookkeeping insert were not atomic, so a crash part-way could leave the schema half-applied while the migration was recorded as complete. Now they commit together.Safe to batch: every statement is
CREATE ... IF NOT EXISTS(tables, indexes, and one fts5 virtual table). No PRAGMAs, no triggers.2. Give the package a 30s
testTimeout/hookTimeout. These suites drive a real file-backed SQLite database through@libsql/client, so wall time is set by the host filesystem rather than by our code. The 5s default is tuned for CPU-bound unit tests. A slow disk should report as slow, not as a spurious failure.Verification
vitest run srcinpackages/openontology— 148 passed, five consecutive runs, no flake.vitest run contractinapps/logicsrc-web— 71 passed, including the 25 ontology-API tests that are the only downstream consumer ofcreateLibsqlStore.tscbuild clean for the package and its dependents.Timeout #2 is the safety net; the batching in #1 is the part that actually reduces the variance.
🤖 Generated with Claude Code