[Linux-cluster] manual fencing -- issues

[gordan at bobich.net]

My understanding is that fence device selection is done by ordering, but I 
could be wrong.

Gordan

On Tue, 29 Jan 2008, Terry wrote:

> Hello,
>
> I know that manual fencing is not supported.  However, I don't have
> anything else with the hardware I am testing at this time.  When (if)
> we go live with this, I'll be using Dell DRACs to handle the fencing.
> At any rate, I am trying manual fencing to test proof of concept.  I
> am using RHEL5.1 with Conga to configure it.  My config is at the
> bottom of this email.   Shouldn't there be a nodename paramater in
> this block:
>                        <fence>
>                                <method name="1">
>                                        <device name="fence00"
> nodename="dss01.foobar.local"/>
>                                </method>
>                        </fence>
>
> ??  If so, why isn't Conga configuring this correctly?  If not, why am
> I getting this error during a crash of the node with the service on
> it:
> Jan 29 10:56:08 dss00 fenced[2965]: agent "fence_manual" reports:
> failed: fence_manual no node name
> Jan 29 10:56:08 dss00 fenced[2965]: fence "dss01.foobar.local" failed
>
> Thanks for any ideas.
>
>
>
>
>
> <?xml version="1.0"?>
> <cluster alias="cluster00" config_version="5" name="cluster00">
>        <fence_daemon clean_start="0" post_fail_delay="0" post_join_delay="3"/>
>        <clusternodes>
>                <clusternode name="dss01.foobar.local" nodeid="1" votes="1">
>                        <fence>
>                                <method name="1">
>                                        <device name="fence01"/>
>                                </method>
>                        </fence>
>                </clusternode>
>                <clusternode name="dss00.foobar.local" nodeid="2" votes="1">
>                        <fence>
>                                <method name="1">
>                                        <device name="fence00"/>
>                                </method>
>                        </fence>
>                </clusternode>
>        </clusternodes>
>        <cman expected_votes="1" two_node="1"/>
>        <fencedevices>
>                <fencedevice agent="fence_manual" name="fence00"/>
>                <fencedevice agent="fence_manual" name="fence01"/>
>        </fencedevices>
>        <rm>
>                <failoverdomains>
>                        <failoverdomain name="failover00" ordered="0"
> restricted="1">
>                                <failoverdomainnode
> name="dss01.foobar.local" priority="1"/>
>                                <failoverdomainnode
> name="dss00.foobar.local" priority="1"/>
>                        </failoverdomain>
>                </failoverdomains>
>                <resources/>
>                <service autostart="1" domain="failover00"
> exclusive="0" name="nfs" recovery="relocate">
>                        <ip address="192.168.100.135" monitor_link="1"/>
>                        <fs device="/dev/sdb1" force_fsck="0"
> force_unmount="1" fsid="8849" fstype="ext3" mountpoint="/data00"
> name="data00" self_fence="1"/>
>                        <nfsexport name="data00_nfs"/>
>                </service>
>        </rm>
> </cluster>
>
> --
> Linux-cluster mailing list
> Linux-cluster at redhat.com
> https://www.redhat.com/mailman/listinfo/linux-cluster
>