Tuning Red Hat Cluster Manager

view manual entry for clumembd
# man clumembd

Parameter tuning for inproper default configuration of Red Hat Cluster Manager
clumembd%broadcast_primary_only
If set, this causes the cluster to only send broadcast packets on the NIC with the IP matching the member name and the loopback device. This is only used when using broadcast heartbeat mode, and is ignored when multicast mode is used.
Current value : 0
New value : 1

cluster%msgsvc_noarp
If set to "yes", the cluster will no longer use an internal ARP table for connection checks. If you are seeing "Not in subnet" messages, try changing this option to "yes". Changing this requires a full restart of the cluster software on all nodes to take effect.
Current value : no
New value : yes

clumembd%adaptive
Cluster membership daemon will allow a grace period for over-loaded members based on the previous amount of time between heartbeat packets
Current value : no
New value : yes

clumembd%interval
Heartbeat interval, in microseconds (default=750000).
Controls the speed by which heartbeat packets are sent.
Failure detection, in seconds = (interval * tko_count / 1000000)
Current value : 330000
New value : 2000000

clumembd%tko_count
Heartbeat declaration count (default=20).
Controls how many heartbeat intervals must elapse without a heartbeat from a given member before being declared out.
Failure detection, in seconds = (interval * tko_count / 1000000)
Current value : 15
New value : 15

clumembd%rtp
Real-Time Priority (0 to 100; default = disabled) of the cluster membership daemon. By default, the membership daemon does not run in real-time priority. You can enable this if you are having problems on heavily loaded systems.
This value should be greater than or equal to cluquorumd%rtp.
Current value : --disabled--
New value : 10

cluquorumd%rtp
Real-Time Priority (0 to 100; default=disabled) of the cluster quorum daemon. By default, the quorum daemon does not run in real-time priority. You can enable this if you are having problems on heavily loaded systems.
This value should be less than or equal to clumembd%rtp.
Current value : --disabled--
New value : 10

shutdown clumanager daemon
# /sbin/service clumanager stop

Tuning cluster parameter :
# cludb -p clumembd%broadcast_primary_only 1
# cludb -p cluster%msgsvc_noarp yes
# cludb -p clumembd%adaptive yes
# cludb -p clumembd%interval 2000000
# cludb -p clumembd%tko_count 15
# cludb -p clumembd%rtp 10
# cludb -p cluquorumd%rtp 10

start clumanager daemon
# /sbin/service clumanager start

cluster configuration can be view at
# more /etc/cluster.xml

Thanks to Pak Firman for gimme a brave to do this things, eventhough it deployed on live systems.