[Gluster-devel] GlusterFS mounted by FUSE BUG

[Kviw John]

hi,
    1. In the FUSE kernel, it will cache the inode and dentry for higher performance. And setting the longer timeout about attribute and entry will guarantee cache fressness. I have read the fuse kernel code, it did exist and enable.
    2. GlusterFS client mounted by "mount -t glusterfs -o attribute-timeout=3600 -o entry-timeout=3600 server1:/test /mnt/glusterfs" didn't cache the inode and dentry, in other words, the fuse kernel metadata cache didn't cache the inode and dentry from my testing.
    3. The inode and dentry were cached in the server OS cache. How to explain the phenomenon or how to enable the fuse kernel to cache inode and dentry in the client rather than the server?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.gluster.org/pipermail/gluster-devel/attachments/20160702/898d42d9/attachment.html>