Only restore Yarn caches on exact key hits #26133
Merged
Conversation
This file contains 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
When we restore a yarn cache that's not specific to a lockfile entry, we might restore packages that are no longer used. When we then run yarn install, we potentially add new packages to the cache. However, yarn is not clearing the unused packages from the global cache. So when we then save the cache we now retain unused packages in the global cache. This means that the global cache grows indefinitely. Restoring the cache isn't free so CI install times will degrade over time. Either we periodically prune the cache, or just not restore anything unless we have an exact hit. The chosen tradeoff depends on the relation of commits changing deps to commits not changing deps.
facebook-github-bot
added
CLA Signed
React Core Team
github-actions bot
pushed a commit
that referenced
this pull request
Feb 9, 2023
## Summary [Current Yarn cache size: 555MB](https://app.circleci.com/pipelines/github/facebook/react/38163/workflows/70d0149e-b0bc-44e8-b8c9-e5c744cab89b/jobs/625334?invite=true#step-102-2) [Used Yarn cache size: 344MB](https://app.circleci.com/pipelines/github/facebook/react/38166/workflows/4825d444-1426-4321-b95b-c540e6cdc6d7/jobs/625354?invite=true#step-104-5) When we restore a global Yarn cache that's not specific to a lockfile entry (i.e. a fallback cache), we might restore packages that are no longer used. When we then run yarn install, we potentially add new packages to the cache. For example: 1. we bump a package version 2. lockfile changes 3. cache restore misses for exact key 4. cache restore hits a prefix (fallback) containing the older version, 5. yarn install adds the new version to the cache Yarn is not clearing the unused packages from the global cache. So when we then save the cache we now retain the old and new version of a package in the global cache even though the old version is no longer used. This means that the global cache grows indefinitely. Restoring the cache isn't free so CI install times will degrade over time. Either we 1. periodically prune the cache 2. just not restore anything unless we have an exact hit. The chosen tradeoff depends on the relation of commits changing deps to commits not changing deps. From my experience, we change deps rarely so I opted to only restore the cache on exact hits. ## How did you test this change? - run on `main` has 555MB of Yarn cache: https://app.circleci.com/pipelines/github/facebook/react/38163/workflows/70d0149e-b0bc-44e8-b8c9-e5c744cab89b/jobs/625334?invite=true#step-102-2 - run on this branch only has 334MB of Yarn cache: https://app.circleci.com/pipelines/github/facebook/react/38166/workflows/4825d444-1426-4321-b95b-c540e6cdc6d7/jobs/625354?invite=true#step-104-5 DiffTrain build for [c0b0b3a](c0b0b3a) [View git log for this commit](https://github.com/facebook/react/commits/c0b0b3a9f80fd57e882859afd95c2f08599442ba)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Summary
Current Yarn cache size: 555MB
Used Yarn cache size: 344MB
When we restore a global Yarn cache that's not specific to a lockfile entry (i.e. a fallback cache), we might restore packages that are no longer used. When we then run yarn install, we potentially add new packages to the cache.
For example:
Yarn is not clearing the unused packages from the global cache. So when we then save the cache we now retain the old and new version of a package in the global cache even though the old version is no longer used.
This means that the global cache grows indefinitely. Restoring the cache isn't free so CI install times will degrade over time.
Either we
The chosen tradeoff depends on the
relation of commits changing deps to commits not changing deps.
From my experience, we change deps rarely so I opted to only restore the cache on exact hits.
How did you test this change?
mainhas 555MB of Yarn cache: https://app.circleci.com/pipelines/github/facebook/react/38163/workflows/70d0149e-b0bc-44e8-b8c9-e5c744cab89b/jobs/625334?invite=true#step-102-2