Description of problem: Following situation is based on a configuration where FUSE mount of a GlusterFS volume is shared via Samba with or without glusterfs_fuse VFS module. With the change[1] to convert error code ENOENT to ESTALE in GlusterFS FUSE bridge layer(for certain operations) Samba's get_real_filename logic which is based on virtual xattrs(in GlusterFS) fails to perform resulting in denial of basic file operations to SMB clients. There are two aspects to be considered - (a) Unable to create or rename files/directories anywhere inside GlusterFS volume shares: This is seen when vfs_glusterfs_fuse module is loaded in Samba while re-exporting FUSE mount of a GlusterFS volume share (b) Reduced performance during creation or renaming of files/directories: Fixing (a) via [2] in GlusterFS leaves us with a performance degradation which would then need a fix from Samba side to correct itself in VFS modules. The current bug report is to track changes for (b) Version-Release number of selected component (if applicable): samba-4.9.8-105 glusterfs-6.0-6 and glusterfs-3.12.2-47.2 How reproducible: Always Steps to Reproduce(error during creation/renaming of files and directories): 1. Create a basic distribute-replicate volume(use glusterfs-6.0-6) 2. Have a Samba-CTDB setup with a share re-exporting FUSE mount of volume created via step 1 with vfs_glusterfs_fuse loaded with "vfs objects = glusterfs_fuse" in smb.conf 3. Try creating a file/directory anywhere inside share Steps to Reproduce(performance issue): 1. Create a basic distribute replicate volume(use glusterfs-3.12.2-47.2) 2. Have a Samba-CTDB setup with a share using vfs_glusterfs module as "vfs objects = glusterfs" in smb.conf 3. Measure the numbers for creating large number of files 4. Update glusterfs to a 6.0-x version containing fix for [2] 5. record new numbers by repeating step 3 6. Compare numbers from step 3 and step 5 Actual results: SMB clients are presented with NT_STATUS_ACCESS_DENIED while trying to create or rename file/directory. And with updated glusterfs performance got reduced. Expected results: SMB clients should be able to create or rename files/directories without any errors and performance should not be impacted. [1] https://code.engineering.redhat.com/gerrit/gitweb?p=rhs-glusterfs.git;a=commit;h=59629f1da9dca670d5dcc6425f7f89b3e96b46bf [2] https://bugzilla.redhat.com/show_bug.cgi?id=1724089
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHSA-2019:3253