why `rsync --sparse` create much larger file than the source?

Wang

10/8/23, 6:21 PM

I try to rsync between 2 local disks, thus there is no compression. the sparse file has apparent size of 50GB, actual size of 16GB. I tried rsync -a -P --sparse, it did generate sparse file on destination at the end. However, it is not as sparse as the source file. The generated file is 27GB instead of 16GB. I can use fallocate -d <filename> to shrink it more. It became around 18GB. However, this operation takes quite long time. In other hand, cp -a --sparse can create identical file.

Also the speed is as slow as the rsync command without --sparse. This is because:

As currently designed, the --sparse option only affects writing of the destination file, not the transfer of the data

However, I am more interested in the remote size. I wonder why rsync did not replicate the exact same size of the source file as cp --sparse=always?

This even happens when the destination is empty. So i guess this has nothing to do with --inplace option.

rsync  version 3.1.3  protocol version 31

according to @anx, I did some test with --block-size=4096 which yield same file size on remote.

747

0 + 4

rsync

sparse-files

anx

10/8/23, 7:11 PM

Is it gone with smaller --block-size? (I wonder if it is just the overhead from rsync not punching unaligned holes)

anx

10/8/23, 7:12 PM

Does this answer your question? [Sparse files significantly larger (but still sparse) after copying over a network](https://serverfault.com/questions/707786/sparse-files-significantly-larger-but-still-sparse-after-copying-over-a-networ)

Wang

10/8/23, 7:51 PM

@anx this does not answer my question at all. I am ok to use `--sparse` every time, and I am using backup. thus I do not use `--inplace`. My concern is more about the size of the sparse file on remote.

Wang

10/8/23, 7:53 PM

@anx check my update on block size

Elon Musk

I sit in a Tesla and translated this thread with Ai:

EN: why `rsync --sparse` create much larger file than the source?

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.