Hello !!!

I build Redhat cluster suite 5.3 with latest cman,openais,rgmanager rpms.
The hardware is dell r610 with fence_drac5 .

Lvm is configured as HA-LVM.

The cluster is working properly , but customer asked for some features which i Hope someone can help me define them.

First :

1.if there is a network failer on one system , Customer ask for 30 sec before fencing The failing system ,& if network is up again – not to fence the system at all.

2.define service which become online only on one system ( on startup ) , but can failover the other cluster node .( no failback )
3.can I define bonding ( mode1 –active/passive) with failback option ??

I attached the cluster.conf file

<?xml version="1.0"?>
<cluster alias="bicluster" config_version="70" name="bicluster">
<quorumd device="/dev/mapper/EmcQdisk" interval="1" label="qdisk" min_score="1" tko="23" votes="1">
<heuristic interval="2" program="ping -c1 -w1 140.4.0.1" score="1"/>
</quorumd>
<fence_daemon clean_start="0" post_fail_delay="0" post_join_delay="3"/>
<clusternodes>
<clusternode name="ubierp1" nodeid="1" votes="1">
<fence>
<method name="1">
<device module_name="" name="ubierp1-drac5"/>
</method>
</fence>
</clusternode>
<clusternode name="ubierp2" nodeid="2" votes="1">
<fence>
<method name="1">
<device module_name="" name="ubierp2-drac5"/>
</method>
</fence>
</clusternode>
</clusternodes>
<cman expected_votes="3" two_node="0"/>
<fencedevices>
<fencedevice agent="fence_drac5" cmd_prompt="admin1-" ipaddr="140.4.20.136" login="root" name="ubierp1-drac5" passwd="eds123"/>
<fencedevice agent="fence_drac5" cmd_prompt="admin1-" ipaddr="140.4.20.137" login="root" name="ubierp2-drac5" passwd="eds123"/>
</fencedevices>
<rm>
<failoverdomains>
<failoverdomain name="etldom" nofailback="1" ordered="1" restricted="0">
<failoverdomainnode name="ubierp1" priority="1"/>
<failoverdomainnode name="ubierp2" priority="2"/>
</failoverdomain>
<failoverdomain name="dwhdom" nofailback="1" ordered="1" restricted="0">
<failoverdomainnode name="ubierp2" priority="1"/>
<failoverdomainnode name="ubierp1" priority="2"/>
</failoverdomain>
</failoverdomains>
<resources>
<ip address="140.4.20.132" monitor_link="1"/>
<ip address="140.4.20.135" monitor_link="1"/>
<lvm lv_name="" name="etl-vg" vg_name="vgetl"/>
<lvm lv_name="" name="dwh-vg" vg_name="vgdwh"/>
<fs device="/dev/mapper/vgetl-etlinfo" force_fsck="0" force_unmount="1" fsid="34796" fstype="ext3" mountpoint="/etl/info" name="etlinfo" options="" self_fence="0"/>
<fs device="/dev/mapper/vgetl-etldata" force_fsck="0" force_unmount="1" fsid="17880" fstype="ext3" mountpoint="/etl/data" name="etldata" options="" self_fence="0"/>
<fs device="/dev/mapper/vgdwh-dwhoradat" force_fsck="0" force_unmount="1" fsid="19572" fstype="ext3" mountpoint="/dwh/data" name="dwhoradat" options="" self_fence="0"/>
<fs device="/dev/mapper/vgdwh-dwhias" force_fsck="0" force_unmount="1" fsid="34067" fstype="ext3" mountpoint="/dwh/ias" name="dwhias" options="" self_fence="0"/>
<fs device="/dev/mapper/vgdwh-dwhbiee" force_fsck="0" force_unmount="1" fsid="29551" fstype="ext3" mountpoint="/dwh/obiee" name="dwhbiee" options="" self_fence="0"/>
</resources>
<service autostart="1" domain="dwhdom" exclusive="1" name="dwhsrv" recovery="relocate">
<ip ref="140.4.20.135"/>
<lvm ref="dwh-vg"/>
<fs ref="dwhoradat"/>
<fs ref="dwhias"/>
<fs ref="dwhbiee"/>
</service>
<service autostart="1" domain="etldom" exclusive="1" name="etlsrv" recovery="relocate">
<ip ref="140.4.20.132"/>
<lvm ref="etl-vg"/>
<fs ref="etlinfo"/>
<fs ref="etldata"/>
</service>
</rm>
</cluster>



Thanks in advance

Shalom.

Hello,
This is a timeout setting thing.
If you want to give the cluster 30secs before declaring a member as dead (this is how I understand your request) you need to change the token timeout in millisecs for the openais totem protocol as follows:
But bare in mind the timeouts in cluster are a very delicate thing.
This should be possible with the etldom (failoverdomain) already specified. To be on the safe side set restricted="1".
I don't recall exactly but I think active/passive bonding does autofailback. If not have a look at here
http://ignum.dl.sourceforge.net/proj...07/bonding.txt
or reactivate the old nic when it comes back as described in the docu above.

Hope this helps.