[Gluster-devel] HA translator issues

[Melvin Wong]

Hi,

I've tried setting up the HA translator using glusterfs-2.0.0rc1 and
fuse-2.7.4glfs11. I have a 2 clients connecting to the 2
glusterfs-servers (lb1 and lb2) which run server-side AFR between
themselves. To test if the HA works, I shutdown the lb1. I could see
both the clients switching to lb2 in the log (fuse-bridge: inode num of
/globalshare changed 54394886 -> 30605318) so it works fine. However
when I restarted lb1 and try to shutdown lb2, 1 of the clients report
"Transport endpoint is not connected" and the logs are as below. Can
someone advise if HA translator is currently stable?

 

 

  1: #############################################

  2: ##  GlusterFS Client Volume Specification  ##

  3: #############################################

  4:

  5: # the exported volume to mount                    # required!

  6: volume cluster

  7:   type protocol/client

  8:   option transport-type tcp/client

10:   option remote-host lb1.world.net

 11:   option remote-subvolume gfs                     # exported volume

 12:   option transport-timeout 5                     # value in
seconds, should be relatively low

 13: end-volume

 14:

 15: # the exported volume to mount                    # required!

 16: volume cluster2

 17:   type protocol/client

 18:   option transport-type tcp/client

20:   option remote-host lb2.world.net

 21:   option remote-subvolume gfs                     # exported volume

 22:   option transport-timeout 5                     # value in
seconds, should be relatively low

 23: end-volume

 24:

 25: volume glusterfs-ha

 26:       type cluster/ha

 27:       subvolumes cluster cluster2

 28: end-volume

 29:

 30:

 31:

+-----

2009-02-05 18:08:01 E [socket.c:708:socket_connect_finish] cluster2:
connection failed (Connection refused)

2009-02-05 18:08:06 W [fuse-bridge.c:304:need_fresh_lookup] fuse-bridge:
inode num of /globalshare changed 54394886 -> 30605318

2009-02-05 18:09:25 E [socket.c:104:__socket_rwv] cluster: readv failed
(Connection reset by peer)

2009-02-05 18:09:25 E [socket.c:566:socket_proto_state_machine] cluster:
socket read failed (Connection reset by peer) in state 1
(192.168.89.151:6996)

2009-02-05 18:09:25 E [saved-frames.c:148:saved_frames_unwind] cluster:
forced unwinding frame type(1) op(READ)

 

Regards,

melvin

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-devel/attachments/20090205/aa8e428a/attachment-0003.html>