Question about Heart Beat lan swtiching in case of failure

[johnsonpk]

Hi ,

I would like know in depth ,how service guard handling HB lan failure for following configuration

1)
Lets say in my cluster I have
One dedicated HB -- lan0
one HB for data+hb- lan1
one standby-------- lan2
all the NIC are connected to a single bridged Net

for the above config , what will happen in case my dedicated HB (ie lan0) fails, will the HB traffic flow through lan1? or service gurad will use lan2 as the standby for lan0?


Also below is from my cluster config.
>>>>From Cmgetconf
------------------------------------

NODE_NAME NODE1

NETWORK_INTERFACE lan0

STATIONARY_IP 10.10.1.120

NETWORK_INTERFACE lan1

HEARTBEAT_IP 192.168.250.3

NETWORK_INTERFACE lan2

NETWORK_INTERFACE lan3

I was wondering after seeing the out put from cmviewconf as the standby lans are showing as non-heart beat network even though all of my NIC are on a single bridged Net and the standby lans are supposed to be stand up for both my HB and stationary lan as per the configuration

From cmviewconf
---------------------------------------


Cluster Node information:



Node ID 1:

Node name: Node1

first lock pv name: /dev/dsk/c2t0d2

first lock disk interface type: fcd_vbus



Network ID 1:

ppa: 0

old_ppa: 0

mac addr: 0x001f29bcfafc

hardware path: 0/1/1/0

network interface name: lan0



IPv4 Information:

subnet: 10.10.1.0

subnet mask: 255.255.255.0

ip address: 10.10.1.120



IPv6 Information:



flags: 4 (Non-Heartbeat Network)

bridged net ID: 1



Network ID 4:

ppa: 1

old_ppa: 0

mac addr: 0x001f29bcfafd

hardware path: 0/1/1/1

network interface name: lan1



IPv4 Information:

subnet: 192.168.250.0

subnet mask: 255.255.255.0

ip address: 192.168.250.3



IPv6 Information:



flags: 5 (Heartbeat Network)

bridged net ID: 1


Network ID 2:

ppa: 2

old_ppa: 0

mac addr: 0x001f29bcface

hardware path: 0/2/2/0

network interface name: lan2



IPv4 Information:

subnet: 0.0.0.0

subnet mask: 0.0.0.0

ip address: 0.0.0.0



IPv6 Information:



flags: 2 (Non-Heartbeat Network)

bridged net ID: 1



Network ID 3:

ppa: 3

old_ppa: 0

mac addr: 0x001f29bcfacf

hardware path: 0/2/2/1

network interface name: lan3



IPv4 Information:

subnet: 0.0.0.0

subnet mask: 0.0.0.0

ip address: 0.0.0.0



IPv6 Information:



flags: 2 (Non-Heartbeat Network)

bridged net ID: 1



Can any one help me to understand why the cmviewconf showing the standby lan as non heart beat network? Is it a bug in the cmgetconf? In case of my HB lan failure, will my cluster go for a toss and reform to a single node cluster?

Appreciate you inputs

Thanks
Johnson

[sujit kumar singh]

Hi Johnson,

NODE_NAME NODE1

NETWORK_INTERFACE lan0

STATIONARY_IP 10.10.1.120

NETWORK_INTERFACE lan1

HEARTBEAT_IP 192.168.250.3

NETWORK_INTERFACE lan2

NETWORK_INTERFACE lan3

Means You have specified lan0 an Stationary IP and that is not heartbeat IP leading you to get a non-heartbeat network for the configuration as viewed in the cmviewconf.


Please change the lan0 as in the Cluster configuration to be HEARTBEAT IP followed by cmcheckconf and cmapplyconf, then cmviewconf shall show lan0 as heartbeat netwok.

lan2 in cmviewconf shows you as non-heartbeat as this is not designated any role now and is simple standby.


in present config, your heartbeat shall flow through lan1 and if lan1 fails shall failover to lan2 the standby if the physical feasibilty is there.

with the change that i said to mention heartbeat Ip as lan0(along witth lan1 as now mentioned as heartbeat), the heartbeat primarily shall flow from lan0, if lan0 fails shall flow through lan1 , in-case lan0 and lan1 both fail shall flow through the standby lan lan2 if lan2 has the physical feasibility for that.

better that you mention lan0 as well as lan1 both as heartbeat lans.

regards
sujit

[Basheer_2]

Hi john,

If your Heartbeat lan0 is dead, then it uses lan1, because it carries data+heartbeats.

Now if there is too much traffic on lan1 and if the cluster doesn't receive heartbeats from the other node ( it waits for NODE_TIMOUT value in the cluster config file).

if it doesn't recevie a heartbeat within the NODE_TIMEOUT, then it thinks the other node is dead and it tries to form the cluster without that node.

[John Bigg]

You have not configured lan0 as a HB lan. This is confirmed by both the ascii info and cmviewconf output.

Standby lans never show up as HB networks. They will take HBs if a primary HB lan fails. If a non HB lan fails they will not.

You should re-configure lan0 as HEARTBEAT_IP.

If you do this lan0 will carry HB traffic along with lan1. If lan0 (or lan1) then fail there will be a switch to the standby lan3 and then it will carry the HB traffic so you will still have 2 HB subnets.

Hope this is clear.

[johnsonpk]

thanks for you inputs.