[Request]: ext4 sparse file size and interaction with Time Machine

[DrCuff]

Feature or enhancement request details

Hi team,

Eric suggested that I drop a note here. TL;DR, sparse images created by container are somewhat generous per container. snapshot and image files come in at 1.5TB which doesn't get managed well with Apple Time Machine backups over SMB. It's the weirdest edge case, but broke backups and took a while to realize that the files are "unsparsed" as they reach the back up. Info here:

https://www.linkedin.com/feed/update/urn:li:activity:7373442020146823169/

The files in ~/Library/Application Support/com.apple.container/ are the ones that become unsparsed with Time Machine. Current work around is to disable Time Machine for that directory and backup directly with container save image:

Disable TM: https://bsky.app/profile/chromamagic.com/post/3lyvjfg5jec2d

Use Save Image: https://bsky.app/profile/chromamagic.com/post/3lyvlr7l72k2k

I marked this as feature as it's not really a "bug", just a strange interaction between Time Machine and some rather generous ext4 defaults. In the end this may well be a Time Machine thing, rather than a "container thing", it only comes to light when using TM for backup.

Low priority, I may well be the only person on the planet doing this, but it was a weekend head scratcher for sure! :-)

Code of Conduct

• I agree to follow this project's Code of Conduct

[DrCuff]

There's been some chatter about apfs and hfs+ for the Time Machine target, not sure if this is helpful or not, but my set up uses apfs (which ought to keep the files sparse), but it doesn't. TM volume is exported from a Mac mini running Sequoia 15.6.1 using SMB to backup multiple Macs just fine and dandy.

Just checked the mount on the server, and it's apfs and has worked well for ages as you can tell by the name of the ancient disk array. hehe.

jcuff@minisrv COLD % mount | grep Pega
/dev/disk11s1 on /Volumes/Pegasus (apfs, local, nodev, nosuid, journaled, noowners)

So yeah, the exclude is the only good work around right now and to use container save image to get tgz backups which are nice and small and work fine and are picked up by Time Machine.

Thinking about this, it's probably pointless to Time Machine snap these container volumes as they are more than likely in flight and are unlikely to restore nicely, unless there's some magic in apfs/container that I'm not aware of. The container save image should always make it consistent. Probably a "best practice" doc needed for how to properly backup containers (we know Time Machine isn't the way for a variety of reasons) would be fabulous.

Hope this helps!

[DrCuff]

Prior art here also to help join the threads.

Probably can close this one out, pardon the noise.

#119

#497

#404

And one more on ext4 disk size that’s tangentially related to this. Sort of.

#518

[DrCuff]

Closing, we pretty much know what's all going on now. Thanks for the help team!

[jglogan]

@DrCuff Thanks for the report, and thanks also for confirming that your backup target is APFS.