Description of problem: ------------------------ After choosing a particular replica as `replica.split-brain-choice', `ls' or `stat' on a file/directory in metadata split-brain does not report correct results always. For e.g., consider the case of mismatching permissions on both replicas. After choosing one of the replicas (say, client-0) as the choice to examine the file, `ls -l' gives the appropriate permissions. On choosing the other replica, (say client-1), `ls -l' does not show the permissions from client-1, instead shows the same permissions as those on client-0. This behavior, however, is not seen in case of mismatching xattrs on both replicas. Version-Release number of selected component (if applicable): ------------------------------------------------------------- glusterfs-3.7.1-6.el7rhgs.x86_64 How reproducible: ----------------- Always Steps to Reproduce: -------------------- 1. Create a 1x2 volume, start it and fuse mount it. 2. Create a file or directory on the mount point. 3. Kill one replica and change the permissions of the file from the mount point. 4. Bring back the killed brick and kill the other replica. 5. Change permissions of the file from the mount point again. 6. Bring back killed brick and examine the permissions of the file using `ls -l' command after choosing a replica as the replica.split-brain-choice on the mount point - # setfattr -n replica.split-brain-choice -v "new-client-1" testfile Actual results: ---------------- Incorrect permissions are shown after selecting a replica as replica.split-brain-choice. Expected results: ------------------ Correct permissions need to be shown on selecting a particular replica as replica.split-brain-choice. Additional info:
Patch posted upstream for review: http://review.gluster.org/11551
Doc text is edited. Please sign off to be included in Known Issues.
Verified as fixed in glusterfs-3.7.1-14.el7rhgs.x86_64. Tested for a file in metadata split-brain in a distribute-replicate setup and found to be working. In case of a directory in metadata split-brain in a distribute-replicate setup, the `replica.split-brain-status' of the directory reports that it is not in split-brain (BZ #1260779). In case of a pure replica, verified that it is possible to examine the directory in metadata split-brain correctly after choosing appropriate choice as `replica.split-brain-choice'.
Please review and sign-off the edited doc text.
Looks good to me.
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://rhn.redhat.com/errata/RHSA-2015-1845.html