Hi ,
I am trying mds failover, but it didnt work, my client
cannot change the connection to backup mds server. I
have 4 X86-64 servers which include 2 mds servers, 1
ost server, and 1 client, OS is RH AS4, lustre version
is 1.4.6.1 . As descript in mannual , i use linux-ha
to detect two mds server.
This is my steps:
1) generate XML file.
Add mds1 and mds2
lmc -m test.xml --add net --node mds1 --nettype lnet
--nid mds1@tcp1
lmc -m test.xml --add net --node mds2 --nettype lnet
--nid mds2@tcp1
and then config mds
lmc -m test.xml --add mds --node mds1 --failover --mds
mds1 --fstype ldiskfs --dev /dev/sdb --size 0
lmc -m test.xml --add mds --node mds2 --failover --mds
mds1 --fstype ldiskfs --dev /dev/sdb --size 0
2) start lustre in mds, ost, and client.
lconf --node ost1 test.xml
lconf --node mds1 test.xml
lconf --node client test.xml
In client, after lconf finished, dmesg show :
No NID found for mds2_UUID.
Maybe this error is the keypoint???
3) start linux-ha in two mds servers
4) poweroff mds1, and linux-ha start mds2''s lustre
service
5) the client should change it''s connection to
mds2, but it did not, and just wait for mds1. If i
start lustre in mds1, the it ok quickly.
Is anybody has some idea abount it???
THX.
____________________________________________________________________________________
It''s here! Your new message!
Get new email alerts with the free Yahoo! Toolbar.
http://tools.search.yahoo.com/toolbar/features/mail/