abnormally large 'processed' data on small LXC

[veehexx]

until about 30minutes ago, i've been using the PVE built in backup method to do full backups of VM&CT's, to a NFS share.
I'm having a tinker with PBS since my current 3rd party backup software is no longer sufficient for my needs and does not sanely handle offsite datastores.

PBS has been added as a storage point in PVE cluster.

a quick test with 2 of my small LXC's suggests something is wonky with one of them but no idea where to look.
ID 107: 8GB disk, with 1.4GB tar.zst backup size in PVE backups.
ID 112: 8GB disk, with 268MB tar.zst backup side in PVE backups.

i run an on-demand backup and point to the PBS data store so i expect similar results to whats in the PVE backup store.
ID 107: processed 394GB, PBS backup size 405GiB. 15 minutes to backup.
ID 112: PBS backup size 740MiB. 7 seconds backup time

PBS real disk usage is 1.77GB (running dedup, encryption & compression)

ok, fair enough they'll be a difference between a tar.zst and whaetver method PBS is using for ID 112. but whats going on with the 107 LXC - it's magnitudes bigger than it should be?!

I havent yet ran on more VM/CT's to see if it's isolated to 1 CT

 

[veehexx]

not really solved, but since is just a SSH jump host with next to no config, i've created a new CT and backups are of a sane size and complete within seconds.
appears something wonky with the original CT

 

[Impact]

There's not enough information to tell for sure but can you try to trim the CT and see if anything changes?
pct config 107 and lvs if using LVM would be of interest.

 

[fabian]

you likely have a big sparse file in the container, which is read and chunked even though it is mostly empty..

https://bugzilla.proxmox.com/show_bug.cgi?id=3832