too many interfaces on same subnet for MC/SG

[T-Mobile Tech Support]

The situation:
primary subnet is 10.133.81
secondary subnet is 10.133.82
I just changed my hb to the same subnet as the secondary and when I run cmcheckconf on the updated ascii file, it complains that there are
"Too many interfaces on server for subnet 10.133.82.0, lan0 and lan5.
Subnet 10.133.82.0 is tagged for heartbeat but lan5 on server isn't"

I have 4 lan cards now on that subnet:
one for secondary connection and one for failover and
one for hb and one for failover.

What can I do to remedy this or am do I only need one failover card for both secondary and hb?

Pat

[Todd McDaniel_1]

Please provide the cmviewcl -v output so we can see your total setup.

iirc, the hearbeat should be on its own subnet.

Why did you move the hb? If the secondary path fails you won't have a hearbeat and then your cluster will have serious issues.

[Steven E. Protter]

Shalom,

A complaint does not mean "will not work"

Having two heartbeat's on the same subnet raises questions as to how robust the cluster will be in the event of a network failure.

You know your network architecture, so you need to decide whether to head the warning.

I would make sure my two heartbeat networks were on different networks and different network hardware.

SEP

[T-Mobile Tech Support]

cmviewcl -v

CLUSTER STATUS
maf down

NODE STATUS STATE
prousg33 down unknown

Quorum_Server_Status:
NAME STATUS STATE
nqs01.unix.gsm1900.org unknown unknown

Network_Parameters:
INTERFACE STATUS PATH NAME
PRIMARY unknown 0/0/6/1/0 lan6
PRIMARY unknown 1/0/4/1/0 lan12
PRIMARY unknown 0/0/1/1/0/4/0 lan1
PRIMARY unknown 0/0/2/1/0 lan14
STANDBY unknown 1/0/1/1/0/4/0 lan8
STANDBY unknown 0/0/4/1/0 lan5
STANDBY unknown 1/0/2/1/0 lan15
STANDBY unknown 1/0/6/1/0 lan13

NODE STATUS STATE
prousg32 down unknown

Quorum_Server_Status:
NAME STATUS STATE
nqs01.unix.gsm1900.org unknown unknown

Network_Parameters:
INTERFACE STATUS PATH NAME
PRIMARY unknown 0/0/8/1/0 lan2
PRIMARY unknown 1/0/8/1/0 lan5
PRIMARY unknown 0/0/0/1/0 lan0
PRIMARY unknown 1/0/14/1/0 lan7
STANDBY unknown 1/0/0/1/0 lan3
STANDBY unknown 0/0/6/1/0 lan1
STANDBY unknown 0/0/2/1/0 lan6
STANDBY unknown 1/0/6/1/0 lan4

UNOWNED_PACKAGES

PACKAGE STATUS STATE AUTO_RUN NODE
pmaf down halted enabled unowned

Policy_Parameters:
POLICY_NAME CONFIGURED_VALUE
Failover configured_node
Failback manual

Script_Parameters:
ITEM STATUS NODE_NAME NAME
Subnet unknown prousg33 10.133.81.0
Subnet unknown prousg33 10.133.82.0
Subnet unknown prousg33 10.132.88.0
Subnet unknown prousg32 10.133.81.0
Subnet unknown prousg32 10.133.82.0
Subnet unknown prousg32 10.132.88.0

Node_Switching_Parameters:
NODE_TYPE STATUS SWITCHING NAME
Primary down prousg33
Alternate down prousg32

and here is theoutput from the cmcheckconf command. Though it says "unable" will this still work?

[root@prousg33:/etc/cmcluster]\# cmcheckconf -v -C /etc/cmcluster/maf.ascii
Checking cluster file: /etc/cmcluster/maf.ascii
Checking nodes ... Done
Checking existing configuration ... Done
Checking for inconsistencies
Too many interfaces on prousg32 for subnet 10.133.82.0, lan0 and lan5
Subnet 10.133.82.0 is tagged for heartbeats but lan5 on prousg32 isn't.
Too many interfaces on prousg32 for subnet 10.133.82.0, lan5 and lan0
Too many interfaces on prousg33 for subnet 10.133.82.0, lan1 and lan12
Subnet 10.133.82.0 is tagged for heartbeats but lan12 on prousg33 isn't.
Too many interfaces on prousg33 for subnet 10.133.82.0, lan12 and lan1
cmcheckconf: Unable to verify cluster file: /etc/cmcluster/maf.ascii.
[root@prousg33:/etc/cmcluster]\#

[Mike Chisholm · Accepted solution]

The configuration you propose is not supported by Serviceguard. Serviceguard does not allow multiple lan cards to be on the same subnet so for example if you had a netmask of 255.255.255.0 and:
lan0 192.1.1.1
lan1 192.1.1.2

this would be an unsupported configuration.

The reason for this is that it can lead to unpredictable behavior with respect to how packets are routed out of the host by the network stack in the operating system. For example, a client may contact the host on the IP configured on lan0 but the host in replying may reply via lan1 if it appears first in the routing table for that subnet. I am not a networking expert so I cannot say for sure all the details of the network area, but I do know that Serviceguard explicitly checks for this type of configuration and will not allow it.

[T-Mobile Tech Support]

Thanks.
I ended up not using the separate heartbeat network and instead made my stationary IP's into heartbeat IP's.

[T-Mobile Tech Support]

thanks.