[Gluster-users] glusterfsd won't restart on one brick

[Kaushal M]

I think I've found the problem. The problem is not with the brick port, but
instead with the unix domain socket used for communication between glusterd
and glusterfsd.

>From the log you provided,

> [2013-07-29 23:34:41.949089] I [glusterfsd.c:1910:main]
> 0-/usr/sbin/glusterfsd: Started running /usr/sbin/glusterfsd version 3.4.0
> (/usr/sbin/glusterfsd -s ir2 --volfile-id home.ir2.lhome-gluster_home -p
> /var/lib/glusterd/vols/home/run/ir2-lhome-gluster_home.pid -S
> /var/run/5a538b707ce5dbf525ba6d01835863bb.socket --brick-name
> /lhome/gluster_home -l /var/log/glusterfs/bricks/lhome-gluster_home.log
> --xlator-option *-posix.glusterd-uuid=9d2d74bf-9055-47a6-b3df-8c2057ea1dd9
> --brick-port 49157 --xlator-option home-server.listen-port=49157)
> [2013-07-29 23:34:41.952053] I [socket.c:3480:socket_init]
> 0-socket.glusterfsd: SSL support is NOT enabled
> [2013-07-29 23:34:41.952087] I [socket.c:3495:socket_init]
> 0-socket.glusterfsd: using system polling thread
> [2013-07-29 23:34:41.952172] E [socket.c:695:__socket_server_bind]
> 0-socket.glusterfsd: binding to failed: Address already in use

socket.glusterfsd is the transport which is bound to the socket file
specified, which is  /var/run/5a538b707ce5dbf525ba6d01835863bb.socket in
this case as specified by the -S option.
The 'binding to' log should have printed the address:port identifier if
this was a log for a port.

> [2013-07-29 23:34:41.952186] E [socket.c:698:__socket_server_bind]
> 0-socket.glusterfsd: Port is already in use
> [2013-07-29 23:34:41.952195] W [rpcsvc.c:1394:rpcsvc_transport_create]
> 0-rpc-service: listening on transport failed

So this is most likely due the zombie process 23744 still listening on the
unix domain socket. Only one bind can be performed on a unix domain socket.
If another bind is tried we get an EADDRINUSE error.

Can you kill 23744, remove /var/run/5a538b707ce5dbf525ba6d01835863bb.socket
and restart the brick using 'gluster volume start'. This should allow it to
start.

~kaushal
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20130731/fc8a11d8/attachment.html>