Most common disk file systems have a maximum of 2^32 inodes, whereas GlusterFS can have 2^64 as far as I know. GlusterFS seems to replicate the directory structure on distributed volumes on all bricks, unlike files which it puts into only one brick. Does this mean that if there are lots of folders, such as each file in a leaf node of a directory tree, the inode count limit of the underlying disk file systems limits the cluster inode count, and you cannot have more than around 4 billion folders on the volume? Also, is there a way to disable the directory replication? That would solve the problem, at least if it doesn't mean sacrificing performance. -Eetu-
