Hi all,
shortly explain our big problem:
glusterfsverson 2.0.8
xenversion 3.4.1
we have 12 GB image domUs with tap:aio
we have two physikal server with 48TB and 2 physikal server with xen. we
use client-side-replikation and we start the domUs from a LUN on
glusterfs which is sized 700GB.
Now we have to test the replication on the 2 glusterfs server started a
domU on it.domU start and now we deaktivate the network on glusterfs
server 1. the domU is runing further
log message on client:
[2009-11-26 17:01:02] E
[client-protocol.c:457:client_ping_timer_expired] gfs-01-01: Server
192.168.11.11:6997 has not responded in the last 10 seconds, disconnecting.
[2009-11-26 17:01:02] E [saved-frames.c:165:saved_frames_unwind]
gfs-01-01: forced unwinding frame type(1) op(FINODELK)
[2009-11-26 17:01:02] E [saved-frames.c:165:saved_frames_unwind]
gfs-01-01: forced unwinding frame type(2) op(PING)
[2009-11-26 17:01:02] N [client-protocol.c:6438:notify] gfs-01-01:
disconnected
[2009-11-26 17:01:50] E [socket.c:745:socket_connect_finish] gfs-01-01:
connection to 192.168.11.11:6997 failed (No route to host)
[2009-11-26 17:01:50] E [socket.c:745:socket_connect_finish] gfs-01-01:
connection to 192.168.11.11:6997 failed (No route to host)
on domU we try a few test:
1. files create before network goes down
2. create while - files after we deaktivate the network on glusterfs 1
We activate the network on glusterfs server1
log message:
[2009-11-26 17:03:13] N [client-protocol.c:5733:client_setvolume_cbk]
gfs-01-01: Connected to 192.168.11.11:6997, attached to remote volume
'domU-images'.
[2009-11-26 17:03:13] N [client-protocol.c:5733:client_setvolume_cbk]
gfs-01-01: Connected to 192.168.11.11:6997, attached to remote volume
'domU-images'.
3. and create after files
all file were here.
but after we reboot the domU - it cames to the filesystem repair modus -
here we have to repair the filesystem of the domU and reboot again.
after reboot the domU start but is completly destroyed:
- some files corupt
- and files we created before network goes down are here. but no while
and after files!!!
so my question:
what should we do?!?
this scenario must work
i.g.: glusterfs server1 went down and the domU have to run further
is there a spezial parameter in glusterfs config?
Roland Fischer
