Published 2025-01-06.
Time to read: 4 minutes.
This page is part of the
git collection.
I have published 5 articles about the Git large file system (LFS). They are meant to be read in order.
- Git Large File System Overview
- Git LFS Client Installation
- Git LFS Server URLs
- Git LFS Filename Patterns & Tracking
- Git LFS Client Configuration & Commands
- Working With Git LFS
- Evaluation Procedure For Git LFS Servers
7 articles are still in process.
Instructions for typing along are given for Ubuntu and WSL/Ubuntu. If you have a Mac, most of this information should be helpful.
This Page Probably Contains Errors
This page is incomplete and may contain errors. It has been published to allow collaboration with fact-checkers. Do not rely on this information yet.
Git LFS Client Configuration
The Git LFS configuration man page follows:
Output of 'man git-lfs-config'
GIT-LFS-CONFIG(5) GIT-LFS-CONFIG(5)
NAME git-lfs-config - Configuration options for git-lfs
CONFIGURATION FILES git-lfs reads its configuration from any file supported by git config -l, including all per-repository, per-user, and per-system Git configuration files.
Additionally, a small number of settings can be specified in a file called .lfsconfig at the root of the repository; see the "LFSCONFIG" section for more details. This configuration file is useful for setting options such as the LFS URL or access type for all users of a repository, especially when these differ from the default. The .lfsconfig file uses the same format as .gitconfig.
If the .lfsconfig file is missing, the index is checked for a version of the file, and that is used instead. If both are missing, HEAD is checked for the file. If the repository is bare, only HEAD is checked. This order may change for checkouts in the future to better match Git’s behavior.
Settings from Git configuration files override the .lfsconfig file. This allows you to override settings like lfs.url in your local environment without having to modify the .lfsconfig file.
Most options regarding git-lfs are contained in the [lfs] section, meaning they are all named lfs.foo or similar, although occasionally an lfs option can be scoped inside the configuration for a remote.
LIST OF OPTIONS General settings • lfs.url / remote.<remote>.lfsurl
The url used to call the Git LFS remote API. Default blank (derive from clone URL).
• lfs.pushurl / remote.<remote>.lfspushurl
The url used to call the Git LFS remote API when pushing. Default blank (derive from either LFS non-push urls or clone url).
• remote.lfsdefault
The remote used to find the Git LFS remote API. lfs.url and branch.*.remote for the current branch override this setting. If this setting is not specified and there is exactly one remote, that remote is picked; otherwise, the default is origin.
• remote.lfspushdefault
The remote used to find the Git LFS remote API when pushing. lfs.url and branch.*.pushremote for the current branch override this setting. If this setting is not set, remote.pushdefault is used, or if that is not set, the order of selection is used as specified in the remote.lfsdefault above.
• lfs.remote.autodetect
This boolean option enables the remote autodetect feaure within Git LFS. LFS tries to derive the corresponding remote from the commit information and, in case of success, ignores the settings defined by remote.lfsdefault and remote.<remote>.lfsurl.
• lfs.remote.searchall
This boolean option enables Git LFS to search all registered remotes to find LFS data. This is a fallback mechanism executed only if the LFS data could not be found via the ordinary heuristics as described in remote.lfsdefault, remote.<remote>.lfsurl and, if enabled, lfs.remote.autodetect.
• lfs.dialtimeout
Sets the maximum time, in seconds, that the HTTP client will wait to initiate a connection. This does not include the time to send a request and wait for a response. Default: 30 seconds
• lfs.tlstimeout
Sets the maximum time, in seconds, that the HTTP client will wait for a TLS handshake. Default: 30 seconds.
• lfs.activitytimeout / lfs.https://<host>.activitytimeout
Sets the maximum time, in seconds, that the HTTP client will wait for the next tcp read or write. If < 1, no activity timeout is used at all. Default: 30 seconds
• lfs.keepalive
Sets the maximum time, in seconds, for the HTTP client to maintain keepalive connections. Default: 30 minutes.
• lfs.ssh.automultiplex
When using the pure SSH-based protocol, whether to multiplex requests over a single connection when possible. This option requires the use of OpenSSH or a compatible SSH client. Default: false on Windows, otherwise true.
• lfs.ssh.retries
Specifies the number of times Git LFS will attempt to obtain authorization via SSH before aborting. Default: 5.
• core.askpass, GIT_ASKPASS
Given as a program and its arguments, this is invoked when authentication is needed against the LFS API. The contents of stdout are interpreted as the password.
• lfs.cachecredentials
Enables in-memory SSH and Git Credential caching for a single 'git lfs' command. Default: enabled.
• lfs.storage
Allow override LFS storage directory. Non-absolute path is relativized to inside of Git repository directory (usually .git).
Note: you should not run git lfs prune if you have different repositories sharing the same storage directory.
Default: lfs in Git repository directory (usually .git/lfs).
• lfs.largefilewarning
Warn when a file is 4 GiB or larger. Such files will be corrupted when using Windows (unless smudging is disabled) with a Git for Windows version less than 2.34.0 due to a limitation in Git. Default: true if the version is less than 2.34.0, false otherwise.
Upload and download transfer settings These settings control how the upload and download of LFS content occurs.
• lfs.concurrenttransfers
The number of concurrent uploads/downloads. Default 8.
• lfs.basictransfersonly
If set to true, only basic HTTP upload/download transfers will be used, ignoring any more advanced transfers that the client/server may support. This is primarily to work around bugs or incompatibilities.
The git-lfs client supports basic HTTP downloads, resumable HTTP downloads (using Range headers), and resumable uploads via tus.io protocol. Custom transfer methods can be added via lfs.customtransfer (see next section). However setting this value to true limits the client to simple HTTP.
• lfs.tustransfers
If set to true, this enables resumable uploads of LFS objects through the tus.io API. Once this feature is finalized, this setting will be removed, and tus.io uploads will be available for all clients.
• lfs.standalonetransferagent
Allows the specified custom transfer agent to be used directly for transferring files, without asking the server how the transfers should be made. The custom transfer agent has to be defined in a lfs.customtransfer.<name> settings group.
• lfs.customtransfer.<name>.path
lfs.customtransfer.<name> is a settings group which defines a custom transfer hook which allows you to upload/download via an intermediate process, using any mechanism you like (rather than just HTTP). path should point to the process you wish to invoke. The protocol between the git-lfs client and the custom transfer process is documented at https://github.com/git-lfs/git-lfs/blob/main/docs/custom-transfers.md
must be a unique identifier that the LFS server understands. When calling the LFS API the client will include a list of supported transfer types. If the server also supports this named transfer type, it will select it and actions returned from the API will be in relation to that transfer type (may not be traditional URLs for example). Only if the server accepts as a transfer it supports will this custom transfer process be invoked.
• lfs.customtransfer.<name>.args
If the custom transfer process requires any arguments, these can be provided here. This string will be expanded by the shell.
• lfs.customtransfer.<name>.concurrent
If true (the default), git-lfs will invoke the custom transfer process multiple times in parallel, according to lfs.concurrenttransfers, splitting the transfer workload between the processes.
• lfs.customtransfer.<name>.direction
Specifies which direction the custom transfer process supports, either "download", "upload", or "both". The default if unspecified is "both".
• lfs.transfer.maxretries
Specifies how many retries LFS will attempt per OID before marking the transfer as failed. Must be an integer which is at least one. If the value is not an integer, is less than one, or is not given, a value of eight will be used instead.
• lfs.transfer.maxretrydelay
Specifies the maximum time in seconds LFS will wait between each retry attempt. LFS uses exponential backoff for retries, doubling the time between each retry until reaching this limit. If a server requests a delay using the Retry-After header, the header value overrides the exponential delay for that attempt and is not limited by this option.
Must be an integer which is not negative. Use zero to disable delays between retries unless requested by a server. If the value is not an integer, is negative, or is not given, a value of ten will be used instead.
• lfs.transfer.maxverifies
Specifies how many verification requests LFS will attempt per OID before marking the transfer as failed, if the object has a verification action associated with it. Must be an integer which is at least one. If the value is not an integer, is less than one, or is not given, a default value of three will be used instead.
• lfs.transfer.enablehrefrewrite
If set to true, this enables rewriting href of LFS objects using url.*.insteadof/pushinsteadof config. pushinsteadof is used only for uploading, and insteadof is used for downloading and for uploading when pushinsteadof is not set.
• lfs.transfer.batchSize
The number of objects to download/upload sent in a single batch request to the LFS server. Default is 100.
This value should be changed with caution, as it can have a significant impact on the performance of the LFS server and the server is free to return an HTTP 413 status code if this value is too high as the Batch API specification states.
Push settings • lfs.allowincompletepush
When pushing, allow objects to be missing from the local cache without halting a Git push. Default: false.
Fetch settings • lfs.fetchinclude
When fetching, only download objects which match any entry on this comma-separated list of paths/filenames. Wildcard matching is as per gitignore(5). See git-lfs-fetch(1) for examples.
• lfs.fetchexclude
When fetching, do not download objects which match any item on this comma-separated list of paths/filenames. Wildcard matching is as per gitignore(5). See git-lfs-fetch(1) for examples.
• lfs.fetchrecentrefsdays
If non-zero, fetches refs which have commits within N days of the current date. Only local refs are included unless lfs.fetchrecentremoterefs is true. Also used as a basis for pruning old files. The default is 7 days.
• lfs.fetchrecentremoterefs
If true, fetches remote refs (for the remote you’re fetching) as well as local refs in the recent window. This is useful to fetch objects for remote branches you might want to check out later. The default is true; if you set this to false, fetching for those branches will only occur when you either check them out (losing the advantage of fetch --recent), or create a tracking local branch separately then fetch again.
• lfs.fetchrecentcommitsdays
In addition to fetching at refs, also fetches previous changes made within N days of the latest commit on the ref. This is useful if you’re often reviewing recent changes. Also used as a basis for pruning old files. The default is 0 (no previous changes).
• lfs.fetchrecentalways
Always operate as if --recent was included in a git lfs fetch call. Default false.
Prune settings • lfs.pruneoffsetdays
The number of days added to the lfs.fetchrecent* settings to determine what can be pruned. Default is 3 days, i.e. that anything fetched at the very oldest edge of the 'recent window' is eligible for pruning 3 days later.
• lfs.pruneremotetocheck
Set the remote that LFS files must have been pushed to in order for them to be considered eligible for local pruning. Also the remote which is called if --verify-remote is enabled.
• lfs.pruneverifyremotealways
Always run git lfs prune as if --verify-remote was provided.
• lfs.pruneverifyunreachablealways
Always run git lfs prune as if --verify-unreachable was provided.
Extensions • lfs.extension.<name>.<setting>
Git LFS extensions enable the manipulation of files streams during smudge and clean. name groups the settings for a single extension, and the settings are: clean The command which runs when files are added to the index smudge The command which runs when files are written to the working copy ** priority The order of this extension compared to others
Other settings • lfs.<url>.access
Note: this setting is normally set by LFS itself on receiving a 401 response (authentication required), you don’t normally need to set it manually.
If set to "basic" then credentials will be requested before making batch requests to this url, otherwise a public request will initially be attempted.
• lfs.<url>.locksverify
Determines whether locks are checked before Git pushes. This prevents you from pushing changes to files that other users have locked. The Git LFS pre-push hook varies its behavior based on the value of this config key. null - In the absence of a value, Git LFS will attempt the call, and warn if it returns an error. If the response is valid, Git LFS will set the value to true, and will halt the push if the user attempts to update a file locked by another user. If the server returns a 501 Not Implemented response, Git LFS will set the value to false. true - Git LFS will attempt to verify locks, halting the Git push if there are any server issues, or if the user attempts to update a file locked by another user. ** false - Git LFS will completely skip the lock check in the pre-push hook. You should set this if you’re not using File Locking, or your Git server verifies locked files on pushes automatically.
+ Supports URL config lookup as described in: https://git-scm.com/docs/git-config#Documentation/git-config.txt-httplturlgt. To set this value per-host: git config --global lfs.https://github.com/.locksverify [true|false].
• lfs.sshtransfer / lfs.<url>.sshtransfer
Configures whether SSH transfers (the pure SSH protocol) are used. By default (or if the value is set to negotiate), the pure SSH protocol is tried first, and then the older hybrid protocol. If always is used, then only the pure SSH protocol is tried. Similarly, if never is used, then only the hybrid protocol is attempted.
• lfs.<url>.contenttype
Determines whether Git LFS should attempt to detect an appropriate HTTP Content-Type header when uploading using the 'basic' upload adapter. If set to false, the default header of Content-Type: application/octet-stream is chosen instead. Default: 'true'.
• lfs.skipdownloaderrors
Causes Git LFS not to abort the smudge filter when a download error is encountered, which allows actions such as checkout to work when you are unable to download the LFS content. LFS files which could not download will contain pointer content instead.
Note that this will result in git commands which call the smudge filter to report success even in cases when LFS downloads fail, which may affect scripts.
You can also set the environment variable GIT_LFS_SKIP_DOWNLOAD_ERRORS=1 to get the same effect.
• GIT_LFS_PROGRESS
This environment variable causes Git LFS to emit progress updates to an absolute file-path on disk when cleaning, smudging, or fetching.
Progress is reported periodically in the form of a new line being appended to the end of the file. Each new line will take the following format:
<direction> <current>/<total files> <downloaded>/<total> <name>
Each field is described below: direction: The direction of transfer, either "checkout", "download", or "upload". current The index of the currently transferring file. total files The estimated count of all files to be transferred. downloaded The number of bytes already downloaded. total The entire size of the file, in bytes. name The name of the file.
• GIT_LFS_FORCE_PROGRESS lfs.forceprogress
Controls whether Git LFS will suppress progress status when the standard output stream is not attached to a terminal. The default is false which makes Git LFS detect whether stdout is a terminal and suppress progress when it’s not; you can disable this behaviour and force progress status even when standard output stream is not a terminal by setting either variable to 1, 'yes' or 'true'.
• GIT_LFS_SKIP_SMUDGE
Sets whether or not Git LFS will skip attempting to convert pointers of files tracked into their corresponding objects when checked out into a working copy. If 'true', '1', 'on', or similar, Git LFS will skip the smudge process in both git lfs smudge and git lfs filter-process. If unset, or set to 'false', '0', 'off', or similar, Git LFS will smudge files as normal.
• GIT_LFS_SKIP_PUSH
Sets whether or not Git LFS will attempt to upload new Git LFS object in a pre-push hook. If 'true', '1', 'on', or similar, Git LFS will skip the pre-push hook, so no new Git LFS objects will be uploaded. If unset, or set to 'false', '0', 'off', or similar, Git LFS will proceed as normal.
• GIT_LFS_SET_LOCKABLE_READONLY lfs.setlockablereadonly
These settings, the first an environment variable and the second a gitconfig setting, control whether files marked as 'lockable' in git lfs track are made read-only in the working copy when not locked by the current user. The default is true; you can disable this behaviour and have all files writeable by setting either variable to 0, 'no' or 'false'.
• lfs.lockignoredfiles
This setting controls whether Git LFS will set ignored files that match the lockable pattern read only as well as tracked files. The default is false; you can enable this behavior by setting the variable to 1, 'yes', or 'true'.
• lfs.defaulttokenttl
This setting sets a default token TTL when git-lfs-authenticate does not include the TTL in the JSON response but still enforces it.
Note that this is only necessary for larger repositories hosted on LFS servers that don’t include the TTL.
LFSCONFIG The .lfsconfig file in a repository is read and interpreted in the same format as the file stored in .git/config. It allows a subset of keys to be used, including and limited to:
• lfs.allowincompletepush
• lfs.fetchexclude
• lfs.fetchinclude
• lfs.gitprotocol
• lfs.locksverify
• lfs.pushurl
• lfs.skipdownloaderrors
• lfs.url
• lfs.\{*}.access
• remote.{name}.lfsurl
The set of keys allowed in this file is restricted for security reasons.
EXAMPLES • Configure a custom LFS endpoint for your repository:
git config -f .lfsconfig lfs.url https://lfs.example.com/foo/bar/info/lfs
SEE ALSO git-config(1), git-lfs-install(1), gitattributes(5), gitignore(5).
Part of the git-lfs(1) suite.
GIT-LFS-CONFIG(5)
Let's examine the git client LFS configuration so far:
Shell
$ git lfs env git-lfs/3.6.0 (GitHub; linux amd64; go 1.23.1) git version 2.45.2
LocalWorkingDir=/mnt/e/media/songs/test LocalGitDir=/mnt/e/media/songs/test/.git LocalGitStorageDir=/mnt/e/media/songs/test/.git LocalMediaDir=/mnt/e/media/songs/test/.git/lfs/objects LocalReferenceDirs= TempDir=/mnt/e/media/songs/test/.git/lfs/tmp ConcurrentTransfers=8 TusTransfers=false BasicTransfersOnly=false SkipDownloadErrors=false FetchRecentAlways=false FetchRecentRefsDays=7 FetchRecentCommitsDays=0 FetchRecentRefsIncludeRemotes=true PruneOffsetDays=3 PruneVerifyRemoteAlways=false PruneVerifyUnreachableAlways=false PruneRemoteName=origin LfsStorageDir=/mnt/e/media/songs/test/.git/lfs AccessDownload=none AccessUpload=none DownloadTransfers=basic,lfs-standalone-file,ssh UploadTransfers=basic,lfs-standalone-file,ssh GIT_EXEC_PATH=/usr/lib/git-core GIT_MERGE_AUTOEDIT=no git config filter.lfs.process = "git-lfs filter-process" git config filter.lfs.smudge = "git-lfs smudge -- %f" git config filter.lfs.clean = "git-lfs clean -- %f"
LFS Cache
As a git user, you know that the git client maintains a local copy of the git database for each repository. The Git LFS client extension does a similar thing, for large files. The LFS stores these files in an LFS cache.
By default, there are at least 2 copies of every large file in LFS-enabled git repositories: one copy in the working tree, and another copy in the Git LFS database.
Shell
$ cd $work/test
$ tree .git/lfs .git/lfs ├── cache │ └── locks │ └── refs │ └── heads │ └── master │ └── verifiable ├── objects │ ├── 32 │ │ └── 63 │ │ └── 326380bc400b0928eca447202e7e6f2b32e547ac10e626849dbb58d1160909fe │ └── 41 │ └── d4 │ └── 41d489370da9b7199d5b6651c7fd16705e0a70be7a7f42bb62f2211079b3a168 └── tmp
12 directories, 3 files
Important Git LFS Commands
After the command-line git client has been enhanced with the Git LFS extension, the following new commands are available:
-
git lfs checkout: updates the working copy with large file content if it has already been downloaded. Does not download any content; seegit-lfs-fetchfor that. -
git lfs env: displays the current Git LFS environment. Helpful for debugging. -
git lfs fetch: Downloads Git LFS objects from the specified remote, or the default remote if not specified. This does not update the working copy; usegit lfs checkoutfor that. -
git lfs fsck: Does a sanity check of the local Git LFS repository. You must rungit lfs fetch --allbefore running this command. -
git lfs install: Performs two functions:- Defines filters in the global git client configuration.
- Installs the Git LFS pre-push hook in the repository that it was called from, if any.
-
git lfs ls-files: only shows committed files. The following example shows a mixture of pointers and objects. Objects are indicated with asterisks, while pointers are indicated by dashes:Shell$ git lfs ls-files 3740843b34 - images/sydney_regatta_gzsim.png 91e96edd71 * vrx_gz/models/coast_waves/meshes/waterlow.dae 3ca6d28451 - vrx_gz/models/dock_block_4x4/mesh/DockCubes_Albedo.png
-
git lfs migrate: See Push and Pull files with Git LFS.-
It seems that the
--aboveoption causes problems because tracking does not consider file size. I would not use this option at this time. -
lfs migrate importcan be extremely slow.
-
It seems that the
-
git lfs prune: releases disk space used by unused files in the Git LFS client cache. This keeps a copy of the latest version of each large file in the Git LFS client cache.
If you want to be aggressive, and do away with the redundant copy in the Git LFS client cache (.git/lfs/objects, not.git/objects), add the--forceoption. For safely, also specify the--verify-remoteand--verify-unreachableoptions, so you do not lose large files from the local git client and have to download them again the next time you need them.Shell$ git lfs prune --force --verify-remote --verify-unreachable
Again, the only objects thatgit-lfs-prunewill remove are locally stored object files, that is, those cached under.git/lfs/objects. The command won't touch anything in the working tree.
The current version ofgit lfs pruneis slow on large repositories; Perhaps a version released in 2025 might improve performance. -
git lfs status: Display paths of Git LFS objects that:- have not been pushed to the Git LFS server. These are large files that would be uploaded by git push.
- have differences between the index file and the current HEAD commit. These are large files that would be committed by git commit.
- have differences between the working tree and the index file. These are files that could be staged using git add.
-
git lfs update: Likegit lfs install, but only performs one function:- Installs the Git LFS pre-push hook in the repository that it was called from.
Git LFS Important Details
I stumbled across this gem of a StackOverflow posting which answered several questions I had been trying in vain to get answers to. It was posted in 2019, 6 years ago, so some details might have changed. Important points:
-
Is this still true?
After running
git lfs track, you must rungit addto refresh the state of files before callinggit lfs statusorgit lfs ls-files. Otherwise, those commands will not report the actual status. -
Some patterns for recursive matching:
git lfs track "*.html"git lfs track "**/*.html"performs identically to the previous patterngit lfs track "directory_name/**/*.html"
- Is this still true? The Git LFS objects not staged for commit section seems misleading as it never displayed any files, even those that should have been tracked by LFS.
- Is this still true? If you attempt to do the previous experiment with a non git-lfs tracked file, you'll also notice that it appears under the Git LFS objects to be committed incorrectly. It is not a Git LFS object. The way you can tell that it's not actually a Git LFS object though is to look at how it ends. If it ends with (Git: 111111111) it won't be committed to LFS.
-
Is this still true?
To avoid the above confusion,
you may want to favor
git lfs ls-filesovergit lfs statusto determine if something is going to be saved in Git LFS or not.
I have published 5 articles about the Git large file system (LFS). They are meant to be read in order.
- Git Large File System Overview
- Git LFS Client Installation
- Git LFS Server URLs
- Git LFS Filename Patterns & Tracking
- Git LFS Client Configuration & Commands
- Working With Git LFS
- Evaluation Procedure For Git LFS Servers
7 articles are still in process.
Instructions for typing along are given for Ubuntu and WSL/Ubuntu. If you have a Mac, most of this information should be helpful.
