[Linux-cluster] XEN VM Cluster
Marcos Ferreira da Silva
marcos at digitaltecnologia.info
Wed Feb 6 19:17:43 UTC 2008
I solved the mount gfs filesystem.
I forgot to run "fence_ack_manual -n vserver1.teste.br".
Now I can mount the gfs partition.
Now I have other problem.
One node appear offline to other node.
When I use command line to migrate online the vm to node 2 it's ok.
And of node 2 to node 1, too.
Appear this in messages file.
Feb 6 17:16:21 vserver1 fenced[5362]: fencing node "vserver2.uniube.br"
Feb 6 17:16:21 vserver1 fenced[5362]: fence "vserver2.uniube.br" failed
VSERVER1
[root at vserver1 ~]# clustat
Member Status: Quorate
Member Name ID Status
------ ---- ---- ------
vserver1.uniube.br 1 Online, Local, rgmanager
vserver2.uniube.br 2 Offline
Service Name Owner (Last) State
------- ---- ----- ------ -----
vm:admin vserver1.uniube.br started
vm:aluno (none) disabled
service:testeAdmin (none) stopped
service:testeAluno (none) stopped
VSERVER2
[root at vserver2 ~]# clustat
Member Status: Quorate
Member Name ID Status
------ ---- ---- ------
vserver1.uniube.br 1 Offline
vserver2.uniube.br 2 Online, Local, rgmanager
Service Name Owner (Last) State
------- ---- ----- ------ -----
vm:admin (none) stopped
vm:aluno (none) disabled
service:testeAdmin vserver2.uniube.br started
service:testeAluno vserver2.uniube.br started
My cluster.conf:
<?xml version="1.0"?>
<cluster alias="cluster1" config_version="24" name="cluster1">
<fence_daemon clean_start="0" post_fail_delay="0"
post_join_delay="3"/>
<fence_xvmd family="ipv4" key_file="/etc/cluster/fence_xvm.key"/>
<clusternodes>
<clusternode name="vserver1.teste.br" nodeid="1" votes="1">
<fence>
<method name="1">
<device domain="admin" name="admin"/>
</method>
</fence>
</clusternode>
<clusternode name="vserver2.teste.br" nodeid="2" votes="1">
<fence>
<method name="1">
<device domain="aluno" name="aluno"/>
</method>
</fence>
</clusternode>
</clusternodes>
<cman expected_votes="1" two_node="1"/>
<fencedevices>
<fencedevice agent="fence_xvm" name="admin"/>
<fencedevice agent="fence_manual" name="manual"/>
</fencedevices>
<rm>
<failoverdomains>
<failoverdomain name="fodAdmin" ordered="1" restricted="1">
<failoverdomainnode name="vserver1.teste.br" priority="2"/>
<failoverdomainnode name="vserver2.teste.br" priority="1"/>
</failoverdomain>
<failoverdomain name="fodAluno" ordered="1" restricted="1">
<failoverdomainnode name="vserver1.teste.br" priority="2"/>
<failoverdomainnode name="vserver2.teste.br" priority="1"/>
</failoverdomain>
</failoverdomains>
<resources>
<clusterfs device="/dev/VGADMIN/LVAdmin" force_unmount="1"
fsid="63078" fstype="gfs" mountpoint="/storage/admin" name="gfsadmin"
options=""/>
<clusterfs device="/dev/VGALUNO/LVAluno" force_unmount="1"
fsid="63078" fstype="gfs" mountpoint="/storage/aluno" name="gfsaluno"
options=""/>
</resources>
<vm autostart="1" domain="fodAdmin" exclusive="0" name="admin"
path="/etc/xen" recovery="relocate"/>
<vm autostart="0" domain="fodAluno" exclusive="0" name="aluno"
path="/etc/xen" recovery="relocate"/>
<service autostart="1" domain="gfsadmin" name="testeAdmin">
<clusterfs ref="gfsadmin"/>
</service>
<service autostart="1" domain="gfsaluno" name="testeAluno">
<clusterfs ref="gfsaluno"/>
</service>
</rm>
</cluster>
Em Qua, 2008-02-06 às 12:07 +0100, jr escreveu:
> > How could configure a partition to share a VM config for two machines?
> > Could you send me your cluster.conf for I compare with I want to do?
>
> no need for my cluster.conf. just use a GFS partition and it will be
> fine. (don't forget to put it into fstab)
>
> >
> > Then I need to have a shared partition to put the VMs config, that will be access by other machines, and a physical (LVM in a storage) to put the real machine.
> > Is it correct?
>
> i don't know what you mean by "real machine", but your guests not only
> need the config, they will also need some storage for their system.
> that's where you need a storage that's connected to your nodes, wether
> it's luns, lvm lvs or image files, no matter. just keep in mind that if
> you are using image files, you need to place them on GFS so that every
> node in your cluster can access them the same.
>
> >
> > When I start a VM in a node 1 it will start in a physical device.
> > If I disconnect the node 1, will the vm migrate to node 2?
> > Will the clients connections lose?
>
> it's just failover, which means that if the cluster sees a problem with
> one of the nodes, the other node will take over it's services, which
> basically means that the vms will be started on the other node.
> that does mean that your clients will get disconnected.
>
> >
> > I'm use a HP Storage and a two servers with multipath with emulex fiber channel.
>
> should be fine.
>
> johannes
>
--
_____________________________
Marcos Ferreira da Silva
DiGital Tecnologia
Uberlândia - MG
(34) 9154-0150 / 3226-2534
More information about the Linux-cluster
mailing list