remote: Counting objects: 3442754, done. remote: Compressing objects: 100% (515633/515633), done. remote: Total 3442754 (delta 2898137), reused 3442437
Recently I encountered the same problem when cloning a git repository on an NFS share with a large git repository (>500MB).
If I ran the git clone
on the server (I.E. not over NFS) the errors went away.
Further testing identified that occasionally data was getting corrupted while being copied to the server via NFSv4.
After much debugging it turned out the issue was related to a buggy network driver corrupting TCP packets when segmentation and rx/tx checksumming was offloaded to the network interface card.
After disabling the segmentation and rx/tx checksum offload on the network interface card (following the instructions in the following blog post: How to solve ssh disconnect packet corrupt problems) the data corruption problems on the NFS share went away, as did my issues with git.