Gluster users - Jul 2016 - GlusterFS client mounted by fuse didn't cache inode/dentry?

hi,
    I have encountered problems that I mounted GlusterFS client as local
filesystem by fuse. By "mount -t glusterfs -o attribute-timeout=86400 -o
entry-timeout=86400 -o direct-io-mode=enable server1:/testvol
/mnt/glusterfs', I mounted GlusterFS to Client1, meanwhile, Client1
didn't exist bricks. In Client1, I excuted a ls shell to list the dentries.
From the "glusterfs(8) - Linux man page", options like
attribute-timeout,entry-timeout will affect the metadata cache in FUSE kernel.
    1, I have monitored the Client1's memory by "free -m" when the
bash shell was excuted, it didn't change any more.
    2, I also have monitored the Server1's memory by "free -m", it
grew about 4G(3 million files).
    3, How to explain the phenomenon? The metadata cache was provided by FUSE
kernel, was there something wrong with my mounting command?
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://www.gluster.org/pipermail/gluster-users/attachments/20160701/0403de00/attachment.html>