Supporting partial cache-hit git pullthroughs
- Dominant language
- Go
- Stars
- 41
- Forks
- 12
- PR merge metrics
- No merged PRs in 30d
Description
I've been experimenting with Cachew a bit, and ran into an unexpected behavior.
Cachew keeps a bare mirror per upstream repo and serves git-upload-pack from it via the git http-backend. On any partial cache miss, cachew fell back to pulling the entire upload-pack request from upstream. I was testing this in a small CI deployment, and when running through the cache it was always doing a full pull when run on a fresh commit.
I was able to get a partial pack pull working by filtering missing objects from the upload-pack request and doing some dynamic serving. Validated it working with a watch on the deployment ingress. Putting together a PR this week if that's a feature the cachew team is open to adding. Cheers!
Contributor guide
Research direction
Start at the git-upload-pack path through git http-backend and trace the partial-cache-miss fallback; the payload names no source file or test, so locate the handler and its existing miss coverage first. Reproduce a fresh-commit pull in the CI deployment, then verify via ingress traffic that only missing objects are fetched rather than the entire upload-pack request.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- git, go
- Domain
- backend, devops
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Quiet
- Clarity
- Mostly clear
- Newbie friendliness
- 45/100