On 23 May 2014 at 10:43, Michael Stone <mst...@debian.org> wrote: > They're the same baseline filesystem, but aufs lets you merge filesystems > and redirect writes so they're logically different. (And, presumably, they > may consume storage differently in other cases than yours and someone may > want to see that via df.) I'd assume that they've got different device IDs > (stat -c %d /mnt/whatever) unlike bind mounts, which are exactly the same > filesystem. (df uses the device IDs to determine which filesystems are > duplicates.)
I'm fairly sure Docker upstream "fixed" this particular problem by changing the way mounts/mount namespaces are handled such that "df" (even "df -a" as root) on the host no longer sees the mounts of all the containers. If someone else could also confirm that their "df" / "df -a" output is clean with a newer Docker (as mine is, on both "overlay" and "aufs" storage backends), then I think we're probably good to close this bug, right? ♥, - Tianon 4096R / B42F 6819 007F 00F8 8E36 4FD4 036A 9C25 BF35 7DD4