I have a backup to a backblaze B2 bucket used just for duplicati. While duplicati reports the size as “Backup: 257.88 GB” on backblaze the size is reported as “Current Size: 277.1 GB”.
Where are the extra 20GB? Could it be that duplicati lost track of some files in B2? Is there a way to clean up the extra space used without uploading everything again?
Note that another backup I have of the same files to local disk is consistent in the reported and actual size used.
Right after posting this it occurred to me that maybe backblaze is listing the size in GB while duplicati is using GiB (but mislabeling it as GB). Is that really the case?
Yes
Great, thanks for the quick and helpful reply!
I searched for my issue here and on Google and couldn’t find these…
No worries, it easily gets buried in 800 issues 
I even cheated. Didn’t have to search cause I remembered creating the issue 
In my opinion any time KB, MB, GB, etc, uses multiples of 1000 it is wrong! It all started by those damn hard drive manufacturers to make their drives appear larger than they really are.
Perhaps I’m just a binary purist. I appreciate that MiB, GiB, TiB, takes away the ambiguity and clearly identifies 2^10 multiples, but it’s a shame that they had to be invented in the first place.
It’s beyond me the entire civilized world can agree to use the metric system, over subpar systems like the imperial system, for everything and still somehow not agree to use the metric system in computers.
The issue is that computers are binary, not decimal. Otherwise I would agree with you. 
The SI system is not something invented by hard drive manufacturers.
The argument that computers internally work binary has nothing to do with how we count bytes. If they would work with hexadecimal numbers internally, kilo would stand for 1099511627776?
The fact that kilo stands for 1000 and not 1024 is a definition and nothing we should exchange opinions about.
If you want to calculate with 1024, that’s not a problem: Use the binary prefixes KiB, MiB, GiB etc.
They are standardizes since 1996. macOS and Linux use it correctly by now. Only Windows is still calculating it wrong and that’s why so many people still believe this must be correct. It is not.
Any wrong use increases the confusion, as this issue shows.