[Linux-cluster] RHCS - concept and understanding of fencing
Chris Kwall
christiankwall-qsa at yahoo.com
Thu Feb 16 10:51:20 UTC 2012
Hi guys
First sorry
if something is unclear, English isn’t my natural language but I still
improving it.
It seems
to be that I’ve a problem in understanding the fencing concept (RHCS 6.2) or I
do something wrong in the configuration. Let me explain the details.
I want
to setup a 2 node Cluster for a simple active/passive Scenario with SAN-Storage
and Quorum Device.
According
to my understanding
For a 2
Node Cluster it’s necessary to set "cman two_node="1"
expected_votes="1".
So every
System has one vote and but one must be available to be quorate.
What is
if I want to have an extra Qurom-Device, provided by SAN Storage.
In this setup
"two_node=1" is not necessary and expected_votes=1 must change to 2.
node1+node2+quorum
= 3. So if node 2 goes down we have 2
votes (node1+quorum disk) and the cluster is still quorate and can still provide
Services. On the other site the quorum disk is available at both system – so every
system has 2 votes in sum 4, that confused me a little bit.
<cluster>
..
<cman expected_votes="3" />
<quorumd label="qdisk" />
..
</cluster>
Q: In
which situation will node1<->node2 fencing them.
· If
the hearbeat-LAN disconnect?
· If
Qurom-Disk went offline?
· If
both goes offline?
In my
understanding of the concept one of both can be down and a node will be fenced
if both goes offline.
Okay –
this was the theory and my understanding of it.
In
praxis my nodes shoot themselves in the head.
Network-Setup
VLAN10 -
192.168.8.0/24 - Backup
VLAN11 -
192.168.100.0/23 - Production
VLAN12 -
192.168.1.0/24 - Cluster / Heartbeat only
VLAN13 -
192.168.10.0/24 – iLO Management
System-Setup
ha-node1
(rhel 6.2)
eth1 - 192.168.8.1
eth2 -
192.168.100.1
eth3 -
192.168.1.1
hp-ilo
192.168.10.1
ha-node2
(rhel 6.2)
eth1 - 192.168.8.2
eth2 -
192.168.100.2
eth3 –
192.168.1.3
hp-ilo
192.168.10.2
/etc/cluster/cluster.conf
(http://pastebin.com/nWhnbk73)
<?xml
version="1.0"?>
<cluster
config_version="4" name="argus">
<fence_daemon/>
<clusternodes>
<clusternode
name="ha-node1" nodeid="1">
<fence>
<method
name="ipmi">
<device action="reboot" name="impifence1"/>
</method>
</fence>
</clusternode>
<clusternode
name="ha-node2" nodeid="2">
<fence>
<method
name="ipmi">
<device
action="reboot" name="impifence2"/>
</method>
</fence>
</clusternode>
</clusternodes>
<cman expected_votes="3"
/>
<fencedevices>
<fencedevice
agent="fence_ipmilan" auth="password"
ipaddr="192.168.10.1" lanplus="1"
login="rhcsfencing" name="impifence1" passwd="*********"/>
<fencedevice
agent="fence_ipmilan" auth="password"
ipaddr="192.168.10.2" lanplus="1"
login="rhcsfencing" name="impifence2"
passwd="*********"/>
</fencedevices>
<rm>
<failoverdomains/>
<resources>
<ip
address="192.168.100.10" monitor_link="on"
sleeptime="10"/>
<ip
address="192.168.8.49" monitor_link="on"
sleeptime="10"/>
<fs
device="/dev/mapper/vg_dataP01-lv_argus" force_fsck="yes"
fstype="ext4" mountpoint="/opt/argus"
name="fsargus"/>
<fs
device="/dev/mapper/vg_dataP01-lv_mysql" force_fsck="yes"
fstype="ext4" mountpoint="/var/lib/mysql"
name="fsmysql"/>
</resources>
<service
autostart="1" name="argus_prod"
recovery="restart">
<ip
ref="192.168.100.10">
<fs
ref="fsmysql">
<script file="/etc/init.d/mysqld"
name="mysqld"/>
</fs>
</ip>
</service>
<service
autostart="1" name="argus_backup"
recovery="restart">
<ip
ref="192.168.8.49"/>
</service>
</rm>
<quorumd label="qdisk" />
</cluster>
/var/log/messages
http://pastebin.com/5G7n56bU
Thx
Chris
More information about the Linux-cluster
mailing list