Hi£¬ All:

I have two physical machines as KVM hosts (clusterA.RHCS and clusterB.RHCS) , an iscsi target set into GFS. 
All I want is a HA Cluster which could migrate all the virtual machines on a node to another when the first node failed into some error status.

So I created a cluster "cluster" using RHCS ,added the two hosts into the cluster . created a fence device .
for every virtual machine on clusterA , create a service ,added a resouce which type is "virtual machine" , and set it's properties.

But when I start the service ,I got these errors in /var/log/message :

[root@clusterA ~]# tail -f /var/log/messages
Nov 24 16:19:29 clusterA libvirtd: Could not find keytab file: /etc/libvirt/krb5.tab: Permission denied
Nov 24 16:19:30 clusterA modclusterd: startup succeeded
Nov 24 16:19:48 clusterA rgmanager[6754]: Executing /etc/init.d/httpd status
Nov 24 16:19:56 clusterA modcluster: Starting service: wuguang-xp on node 
Nov 24 16:19:56 clusterA rgmanager[2198]: Stopping service vm:wuguang-xp
Nov 24 16:19:56 clusterA rgmanager[6850]: Using /etc/libvirt/qemu/wuguang-xp.xml instead of searching /etc/libvirt/qemu
Nov 24 16:19:56 clusterA rgmanager[6906]: xend/libvirtd is dead; cannot stop wuguang-xp
Nov 24 16:19:56 clusterA rgmanager[2198]: stop on vm "wuguang-xp" returned 1 (generic error)
Nov 24 16:19:56 clusterA rgmanager[2198]: Marking vm:wuguang-xp as 'disabled', but some resources may still be allocated!
Nov 24 16:19:56 clusterA rgmanager[2198]: Service vm:wuguang-xp is disabled


But I set the libvirt tcp authority to "none" , here is my libvirtd.conf:


listen_tls = 0


listen_tcp = 1


tcp_port = "16509"
 
auth_tcp = "none"



my cluster.conf is :

<?xml version="1.0"?>
<cluster config_version="83" name="cluster">
<fence_daemon clean_start="0" post_fail_delay="0" post_join_delay="3"/>
<clusternodes>
<clusternode name="clusterA.RHCS" nodeid="1" votes="1">
<fence>
<method name="fence1">
<device name="fence1"/>
</method>
</fence>
</clusternode>
<clusternode name="clusterB.RHCS" nodeid="2" votes="1">
<fence>
<method name="fence2">
<device name="fence2"/>
</method>
</fence>
</clusternode>
</clusternodes>
<cman expected_votes="1" two_node="1"/>
<fencedevices>
<fencedevice agent="fence_scsi" name="fence1" nodename="clusterA.RHCS"/>
<fencedevice agent="fence_scsi" name="fence2" nodename="clusterB.RHCS"/>
</fencedevices>
<rm>
<failoverdomains>
<failoverdomain name="FD" nofailback="0" ordered="1" restricted="1">
<failoverdomainnode name="clusterA.RHCS" priority="1"/>
<failoverdomainnode name="clusterB.RHCS" priority="2"/>
</failoverdomain>
</failoverdomains>
<resources>
<ip address="192.168.6.250/24" monitor_link="on" sleeptime="10"/>
<script file="/etc/init.d/httpd" name="www_script"/>
<clusterfs device="/dev/dm-0" force_unmount="on" fsid="56449" fstype="gfs2" mountpoint="/var/lib/libvirt/images/iscsi" name="gfs2" self_fence="on"/>
</resources>
<service autostart="1" domain="FD" exclusive="0" name="www_service" recovery="relocate">
<ip ref="192.168.6.250/24"/>
<script ref="www_script"/>
<clusterfs ref="gfs2"/>
</service>
<vm autostart="1" domain="FD" exclusive="0" migrate="live" name="wuguang-xp" path="/etc/libvirt/qemu" recovery="relocate" xmlfile="/etc/libvirt/qemu/wuguang-xp.xml"/>
</rm>
</cluster>

What's wrong with my configuration of RHCS or libvirt?
Thanks in advance
 
wade
 

xuteng via foxmail