Microsoft DevOps Engineer Expert AZ-400 Practice Question

Your company hosts a monolithic Git repository in Azure Repos that has grown to 250 GB and contains more than 100 000 files. Developers state that an initial clone takes over an hour and each fetch still transfers hundreds of megabytes. All contributors use Git 2.35 or later on Windows and macOS. You must reduce clone and fetch times without rewriting history or splitting the repository. Which approach should you recommend?

Deploy Scalar on developer workstations and configure sparse-checkout (cone mode) with Git partial clone so that blobs are downloaded only when accessed.
Run git filter-repo on the server to delete historical objects older than one year and force all developers to re-clone the cleaned repository.
Migrate the repository from Git to Team Foundation Version Control (TFVC) to benefit from server-side workspace mapping.
Enable Git LFS for the entire repository, track every file larger than 1 MB, and instruct developers to re-clone using the --no-checkout option.

Report Issue

Answer Description

Scalar installs a service that automatically configures Git partial clone and sparse-checkout (cone mode). With partial clone, only the metadata and tree objects are downloaded initially; file contents (blobs) are fetched on demand, reducing the size and duration of clone and fetch operations. Sparse-checkout further limits the working directory to the paths a developer needs. The server continues to store full history, so no history rewrite or repository migration is required.

A history rewrite with git filter-repo could shrink the repo, but it breaks commit hashes and forces every consumer (including CI pipelines) to migrate, which contradicts the requirement to avoid rewriting history. Switching to TFVC is a migration to a different version-control system rather than an optimization of the existing Git repository. Enabling Git LFS for every file larger than 1 MB would create thousands of pointers and still require an extensive migration; LFS improves binary handling but would not significantly reduce metadata or tree sizes that dominate clone and fetch times for very large repositories.

Ask Bash

Bash is our AI bot, trained to help you pass your exam. AI Generated Content may display inaccurate information, always double-check anything important.