[Gluster-users] Glusterfs self-heal on devices minor/major numbers

Vadim S. Khondar v.khondar at o3.ua
Fri Mar 18 17:55:51 UTC 2011


Hello everyone,

I have a setup with two servers running glusterfs 3.1.2 (Repository
revision: v3.1.1-64-gf2a067c) with fuse init (API version 7.10) on
CentOS 5.5.

Glusterfs is used for replicated storage volume for openvz images.
The case is when I restore image of VPS on one server, device entries
in /dev of VPS are created with correct minor/major numbers. These are
also synced to another server and appear there soon but with wrong
minors/majors.

Here are parts of `ls -la' within /store (this directory acts as storage
backend for gluster volume) on both servers:

srv1:
...
crw-rw-rw-  1 root root 5,   1 Кві 13  2006 console
crw-------  1 root root 1,   6 Apr 13  2006 core
crw-r-----  1 root kmem 1,   2 Apr 13  2006 kmem
crw-------  1 root root 1,  11 Apr 13  2006 kmsg
crw-r-----  1 root kmem 1,   1 Apr 13  2006 mem
crw-rw-rw-  1 root root 1,   3 Apr 13  2006 null
crw-r-----  1 root kmem 1,   4 Apr 13  2006 port
crw-rw-rw-  1 root tty  3, 176 Apr 13  2006 ttya0
crw-rw-rw-  1 root tty  3, 177 Apr 13  2006 ttya1
crw-rw-rw-  1 root tty  3, 178 Apr 13  2006 ttya2
...

srv2:
...
crw-rw-rw-  1 root root 253, 1 Apr 13  2006 console
crw-------  1 root root 253, 1 Apr 13  2006 core
crw-r-----  1 root kmem 253, 1 Apr 13  2006 kmem
crw-------  1 root root 253, 1 Apr 13  2006 kmsg
crw-r-----  1 root kmem 253, 1 Apr 13  2006 mem
crw-rw-rw-  1 root root 253, 1 Apr 13  2006 null
crw-r-----  1 root kmem 253, 1 Apr 13  2006 port
crw-rw-rw-  1 root root 253, 1 Apr 13  2006 ptmx
crw-rw-rw-  1 root root 253, 1 Apr 13  2006 tty
crw-rw-rw-  1 root tty  253, 1 Apr 13  2006 ttya0
crw-rw-rw-  1 root tty  253, 1 Apr 13  2006 ttya1
crw-rw-rw-  1 root tty  253, 1 Apr 13  2006 ttya2
...

As you see, device numbers are screwed up.

I restore VPS not to /store, but to mounted glusterfs volume.
Besides, if I delete everything from wrong /dev and trigger self-heal
with something like `touch -a -c dev/*' device entries are being
recreated with same screwed maj/mins.

I wonder if this is a bug and if there is any solution beside manual fix
on backend storages of gluster volume?

-- 
With best regards,
Vadim




More information about the Gluster-users mailing list