Created attachment 1026857 [details] Required detailed information as per gluster bug reporting guidelines Description of problem: If the first specified volfile-server is not responding while mounting a volume, specified backup servers are *not contacted at all*. Version-Release number of selected component (if applicable): 3.6.3 (unofficial Ubuntu 14.04.2 LTS build based on Debian source packages) How reproducible: *Always* by ensuring that first --volfile-server passed to /usr/sbin/glusterfs is not responding. Steps to Reproduce: 1. Modify a working /etc/fstab entry so that the server referenced on the left is not responding (alternatively, shut down the named server) E.g., in the following, it is sufficient to ensure that 10.0.3.83 is unavailable while 10.0.3.81, 10.0.3.82 are up and running just fine: 10.0.3.83:/gvol00 /mirrored/gluster glusterfs defaults,_netdev,backup-volfile-servers=10.0.3.81:10.0.3.82 0 0 2. Execute "sudo umount /mirrored/gluster; sudo mount /mirrored/gluster" to ensure that the above configuration will take effect; the second command will fail. 3. Counter-example: "Revive" 10.0.3.83 or just permute the IP addresses above so that a working node is contacted first and re-run the previous command which will now succeed. Actual results: The volume can be mounted on the client iff and only iff the first volfile-server is available at that point in time. Expected results: All specified volfile-servers are contacted at least once when trying to mount the volume in question (regardless of an existing fetch-attempts option which does not change the behaviour) Additional info: Please refer to the attached files which contain details, statedumps.
Created attachment 1026858 [details] statedump (node 1/2, host "galera01")
Created attachment 1026859 [details] statedump (node 2/2, host "galera02")
Previous wording might be misleading--it is expected that the volfile-servers are contacted one after another until mounting succeeds, of course.
I had below entry in /etc/fstab > example.com:/demo /mnt/glusterfs glusterfs backup-volfile-servers=tumballi.in:local 0 0 where only 'local' was a proper valid hostname where the volume demo was hosted. It did work when I did `mount /mnt/glusterfs`. It took some time, (~20second in this case) but it worked fine. Considering a lot of time is passed since the bug was open, I would say, you may need to upgrade to higher version to see that it is fixed. (I don't have exact patch link which fixed this).