Partial Clone

As Git repositories grow in size, they can become cumbersome to work with because of the large amount of history that must be downloaded, and the large amount of disk space they require.

Partial clone is a performance optimization that "allows Git to function without having a complete copy of the repository. The goal of this work is to allow Git better handle extremely large repositories."

Git 2.22.0 or later is required.

Filter by file size

Introduced in GitLab 12.10.

Storing large binary files in Git is normally discouraged, because every large file added will be downloaded by everyone who clones or fetches changes thereafter. This is slow, if not a complete obstruction when working from a slow or unreliable internet connection.

Using partial clone with a file size filter solves this problem, by excluding troublesome large files from clones and fetches. When Git encounters a missing file, it will be downloaded on demand.

When cloning a repository, use the --filter=blob:limit=<size> argument. For example, to clone the repository excluding files larger than 1 megabyte:

git clone --filter=blob:limit=1m git@gitlab.com:gitlab-com/www-gitlab-com.git

This would produce the following output:

Cloning into 'www-gitlab-com'...
remote: Enumerating objects: 832467, done.
remote: Counting objects: 100% (832467/832467), done.
remote: Compressing objects: 100% (207226/207226), done.
remote: Total 832467 (delta 585563), reused 826624 (delta 580099), pack-reused 0
Receiving objects: 100% (832467/832467), 2.34 GiB | 5.05 MiB/s, done.
Resolving deltas: 100% (585563/585563), done.
remote: Enumerating objects: 146, done.
remote: Counting objects: 100% (146/146), done.
remote: Compressing objects: 100% (138/138), done.
remote: Total 146 (delta 8), reused 144 (delta 8), pack-reused 0
Receiving objects: 100% (146/146), 471.45 MiB | 4.60 MiB/s, done.
Resolving deltas: 100% (8/8), done.
Updating files: 100% (13008/13008), done.
Filtering content: 100% (3/3), 131.24 MiB | 4.65 MiB/s, done.

The output will be longer because Git will first clone the repository excluding files larger than 1 megabyte, and second download any missing large files needed to checkout the master branch.

When changing branches, Git may need to download more missing files.

Filter by object type

Introduced in GitLab 12.10.

For enormous repositories with millions of files, and long history, it may be helpful to exclude all files and use in combination with sparse-checkout to reduce the size of your working copy.

# Clone the repo excluding all files
$ git clone --filter=blob:none --sparse git@gitlab.com:gitlab-com/www-gitlab-com.git
Cloning into 'www-gitlab-com'...
remote: Enumerating objects: 678296, done.
remote: Counting objects: 100% (678296/678296), done.
remote: Compressing objects: 100% (165915/165915), done.
remote: Total 678296 (delta 472342), reused 673292 (delta 467476), pack-reused 0
Receiving objects: 100% (678296/678296), 81.06 MiB | 5.74 MiB/s, done.
Resolving deltas: 100% (472342/472342), done.
remote: Enumerating objects: 28, done.
remote: Counting objects: 100% (28/28), done.
remote: Compressing objects: 100% (25/25), done.
remote: Total 28 (delta 0), reused 12 (delta 0), pack-reused 0
Receiving objects: 100% (28/28), 140.29 KiB | 341.00 KiB/s, done.
Updating files: 100% (28/28), done.

$ cd www-gitlab-com

$ git sparse-checkout init --cone

$ git sparse-checkout add data
remote: Enumerating objects: 301, done.
remote: Counting objects: 100% (301/301), done.
remote: Compressing objects: 100% (292/292), done.
remote: Total 301 (delta 16), reused 102 (delta 9), pack-reused 0
Receiving objects: 100% (301/301), 1.15 MiB | 608.00 KiB/s, done.
Resolving deltas: 100% (16/16), done.
Updating files: 100% (302/302), done.

For more details, see the Git documentation for sparse-checkout.

Filter by file path

WARNING: Partial Clone using sparse filters is experimental, slow, and will significantly increase Gitaly resource utilization when cloning and fetching.

Deeper integration between Partial Clone and Sparse Checkout is being explored through the --filter=sparse:oid=<blob-ish> filter spec, but this is highly experimental. This mode of filtering uses a format similar to a .gitignore file to specify which files should be included when cloning and fetching.

For more details, see the Git documentation for rev-list-options.

Create a filter spec. For example, consider a monolithic repository with many applications, each in a different subdirectory in the root. Create a file shiny-app/.filterspec using the GitLab web interface:

# Only the paths listed in the file will be downloaded when performing a
# partial clone using `--filter=sparse:oid=shiny-app/.gitfilterspec`

# Explicitly include filterspec needed to configure sparse checkout with
# git config --local core.sparsecheckout true
# git show master:snazzy-app/.gitfilterspec >> .git/info/sparse-checkout
shiny-app/.gitfilterspec

# Shiny App
shiny-app/

# Dependencies
shimmery-app/
shared-component-a/
shared-component-b/

Create a new Git repository and fetch. Support for --filter=sparse:oid using the clone command is incomplete, so we will emulate the clone command by hand, using git init and git fetch. Follow issue tracking support for --filter=sparse:oid for updates.

# Create a new directory for the Git repository
mkdir jumbo-repo && cd jumbo-repo

# Initialize a new Git repository
git init

# Add the remote
git remote add origin <url>

# Enable partial clone support for the remote
git config --local extensions.partialClone origin

# Fetch the filtered set of objects using the filterspec stored on the
# server. WARNING: this step is slow!
git fetch --filter=sparse:oid=master:shiny-app/.gitfilterspec origin

# Optional: observe there are missing objects that we have not fetched
git rev-list --all --quiet --objects --missing=print | wc -l

WARNING: Git integrations with bash, zsh, etc and editors that automatically show Git status information often run git fetch which will fetch the entire repository. You many need to disable or reconfigure these integrations.

Sparse checkout must be enabled and configured to prevent objects from other paths being downloaded automatically when checking out branches. Follow issue proposing automating sparse checkouts for updates.

# Enable sparse checkout
git config --local core.sparsecheckout true

# Configure sparse checkout
git show master:snazzy-app/.gitfilterspec >> .git/info/sparse-checkout

# Checkout master
git checkout master

Remove partial clone filtering

Git repositories with partial clone filtering can have the filtering removed. To remove filtering:

Fetch everything that has been excluded by the filters, to make sure that the repository is complete. If git sparse-checkout was used, use git sparse-checkout disable to disable it. See the disable documentation for more information.

Then do a regular fetch to ensure that the repository is complete. To check if there are missing objects to fetch, and then fetch them, especially when not using git sparse-checkout, the following commands can be used:

# Show missing objects
git rev-list --objects --all --missing=print | grep -e '^\?'

# Show missing objects without a '?' character before them (needs GNU grep)
git rev-list --objects --all --missing=print | grep -oP '^\?\K\w+'

# Fetch missing objects
git fetch origin $(git rev-list --objects --all --missing=print | grep -oP '^\?\K\w+')

# Show number of missing objects
git rev-list --objects --all --missing=print | grep -e '^\?' | wc -l

Repack everything. This can be done using git repack -a -d, for example. This should leave only three files in .git/objects/pack/:
- A pack-<SHA1>.pack file.
- Its corresponding pack-<SHA1>.idx file.
- A pack-<SHA1>.promisor file.
Delete the .promisor file. The above step should have left only one pack-<SHA1>.promisor file, which should be empty and should be deleted.
Remove partial clone configuration. The partial clone-related configuration variables should be removed from Git configuration files. Usually only the following configuration must be removed:
- remote.origin.promisor.
- remote.origin.partialclonefilter.