[JENKINS-75480] Git heavyweight checkout should use partial clone + sparse-checkout

Hi,

We are using multibranch pipelines with "Merging the pull request with the current target branch revision" strategy. We are also using a SCM that do not provide capabilities for lightweight checkout.

As expected, Jenkins triggers a heavyweight checkout for each PR, which stores a full clone of the project for each PR branch.

We are already using some of the usual techniques to reduce bandwidth and storage, including:

• setting sparse-checkout (but the sparse-checkout also applies to clones on agents)


• manually setting reference repositories with a scheduled pull


• scheduled git repack -adl to compress the clones

However, in the end, this doesn't scale.

Initial checkout for retrieving the Jenkinsfile could be vastly improved by updating the heavyweight checkout in the following way, provided that the version of git installed on the controller is recent enough:

• perform a blobless --filter=blob:none --no-checkout clone


• set up a sparse-checkout filter to only read the Jenkinsfile


• perform a checkout and merge to read the Jenkinsfile

note: if the SCM doesn't support partial clone, git reverts the partial clone to a standard clone.

---
Originally reported by tdubouch, imported from: Git heavyweight checkout should use partial clone + sparse-checkout

• status: Open
  
• priority: Minor
  
• component(s): git-client-plugin
  
• resolution: Unresolved
  
• votes: 1
  
• watchers: 2
  
• imported: 20251211-071809
  

Raw content of original issue

Hi,


We are using multibranch pipelines with "Merging the pull request with the current target branch revision" strategy. We are also using a SCM that do not provide capabilities for lightweight checkout.


As expected, Jenkins triggers a heavyweight checkout for each PR, which stores a full clone of the project for each PR branch.


We are already using some of the usual techniques to reduce bandwidth and storage, including:



	
setting sparse-checkout (but the sparse-checkout also applies to clones on agents)

	
manually setting reference repositories with a scheduled pull

	
scheduled git repack -adl to compress the clones



However, in the end, this doesn't scale.


Initial checkout for retrieving the Jenkinsfile could be vastly improved by updating the heavyweight checkout in the following way, provided that the version of git installed on the controller is recent enough:



	
perform a blobless --filter=blob:none --no-checkout clone

	
set up a sparse-checkout filter to only read the Jenkinsfile

	
perform a checkout and merge to read the Jenkinsfile



note: if the SCM doesn't support partial clone, git reverts the partial clone to a standard clone.