perf(isolator): run per-capsule installs sequentially instead of all at once#10457
Open
davidfirst wants to merge 4 commits into
Open
perf(isolator): run per-capsule installs sequentially instead of all at once#10457davidfirst wants to merge 4 commits into
davidfirst wants to merge 4 commits into
Conversation
When rebuilding scope-aspects capsules with nesting enabled, each capsule was installed via its own separate package-manager install. For a cold rebuild of many capsules this means N separate installs that each re-resolve a heavily-overlapping dependency tree and serialize on the pnpm store lock (~6m for 5 env capsules). When more than one capsule needs installing, fall back to a single batched install; capsule isolation is preserved via rootComponentsForCapsules (~1m for the same 5 capsules). The single-capsule incremental path is unchanged.
PR Summary by QodoBatch-install capsules when nesting enabled to avoid N installs Description
Diagram
High-Level Assessment
Files changed (1)
|
Code Review by Qodo
1.
|
![qodo-free-for-open-source-projects[bot]](https://avatars.githubusercontent.com/in/1068000?s=60&v=4){kind=small}
…capsule-install # Conflicts: # scopes/component/isolator/isolator.main.runtime.ts
|
Code review by qodo was updated up to the latest commit dfd76c7 |
zkochan
approved these changes
Jun 26, 2026
Batching multiple scope-aspect capsules into a single install breaks the per-capsule dependency isolation the e2e suite enforces (deps-in-capsules, root-components). Direct-dep placement is fixable via pnpm hoistPattern/ dedupeDirectDeps, but per-capsule peer/version isolation (two versions of the same aspect needing different peers) leaks in a shared install. Per-capsule nesting is required for correctness, so revert to master's behavior.
|
Code review by qodo was updated up to the latest commit 3dba919 |
…at once Each scope-aspect/env capsule is installed in its own package-manager install, previously all fired at once via Promise.all. These installs serialize on the pnpm store lock, so full concurrency mostly thrashes (CPU oversubscription + lock contention) and is actually slower than running them one at a time. Cap the per-capsule install concurrency (1 = sequential): faster wall-clock and far less CPU, with isolation unchanged.
davidfirst
changed the title
|
Code review by qodo was updated up to the latest commit 262ff2c |
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
2 participants
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.
Scope-aspect/env capsules are each installed in their own package-manager install. They were all fired at once via
Promise.all, but these installs serialize on the pnpm store lock, so full concurrency mostly thrashes (CPU oversubscription + lock contention) and ends up slower than running them one at a time.This caps the per-capsule install concurrency to 1 (sequential). On a cold rebuild of 5 env capsules it's both faster and far lighter on CPU:
Promise.all(all at once): ~7m41sLower CPU matters on CI where other jobs share the box. Isolation is unchanged — these are the exact same per-capsule installs, just throttled (kept as a named constant so the value is easy to tune).
Why not batch into one install
The original idea here was to batch all capsules into a single install to skip the repeated resolution. That breaks per-capsule isolation: capsules are loaded and executed at runtime, so each must resolve its own dependency tree (including peers like React) to the exact versions it needs. A single shared install hoists/unifies deps across capsules — caught by the
deps-in-capsulesandroot-componentse2e tests (e.g. two versions of the same aspect resolving the wrong React). Direct-dep placement was fixable with pnpm hoisting flags, but per-capsule peer/version isolation wasn't, so the batching approach was dropped in favor of this safe concurrency change.