On 22.07.2015 09:08, zhaohua.cui(a)majorbio.com wrote:
Jul 22 15:01:27 localhost kernel: LustreError:
1517:0:(ldlm_lib.c:433:client_obd_setup()) can't add initial connection
Jul 22 15:01:27 localhost kernel: LustreError: 1517:0:(obd_config.c:572:class_setup())
setup majorbio-MDT0000-mdc-ffff88007a751000 failed (-2)
Jul 22 15:01:27 localhost kernel: LustreError:
1517:0:(obd_config.c:1591:class_config_llog_handler()) MGC192.168.10.110@tcp: cfg command
failed: rc = -2
Jul 22 15:01:27 localhost kernel: Lustre: cmd=cf003 0:majorbio-MDT0000-mdc
1:majorbio-MDT0000_UUID 2:192.168.10.110@o2ib
Jul 22 15:01:27 localhost kernel: LustreError: 15c-8: MGC192.168.10.110@tcp: The
configuration from log 'majorbio-client' failed (-2). This may be the result of
communication errors between this node and the MGS, a bad configuration, or other errors.
See the syslog for more information.
I'm not an expert on lnet, but I suspect
the above means, the client
tries to use o2ib instead of tcp, which doesn't work from the VM as it
does see the network device as an eth device. Maybe lnet mixes things up
because you use the same IP for both o2ib and tcp. Try to disable o2ib
in the server mgsnode specification, or use 2 IP addresses on the
servers IB interface, and configure one for tcp and one for o2ib.