Service Guard A.11.20 Version erros

[shabash]

The network team provided me 2 ips from same subnet and VLAN.

when i run query,the service guard gives no error but when cmcheckconf then service guard gives error.
lan0 =Ip
lan2 heartbeat
lan1 failover

lan3 configured for backup,not part of cluster.but configured with same subnet as Lan0


error:
Communication for IP at lan0 to lan3 working but not communication occurs on DLPI level.

when i disconnet lan3 the cluster runs fine.

I need to know that can serviceguard allow using same subnet ips configure on 2 lans in system.???

[melvyn burnard]

no, you cannot have two active cards configured on the same subnet.

[Mel Burslan]

Actually, serviceguard or not, having two network interfaces on the same subnet for any given single server is unsupported at best and most of the time flat out doesn't work. Why would you want two network cards on the same subnet anyway ? Increase the throughput ? If so, consider investing in APA software.

[shabash]

Appreciate for the information.

Could any one send link for Hp document regarding this by "highlighting the subnet info line"

we need to communicated this to customer with proper Hp link.

Any Help?

[shabash]

what if that in the system we donot want to include the other lan interface for service guard??

i have tried one below scenerio

For example configure lan0,lan1,lan2 in which lan0 subnet 1
lan2 private H.B local IP(not as lan0)
lan1 standby

when cluster up then assig same subnet of lan0 to lan3 interface.

Then halt cluster and then cmruncl it start gives error like Assertion value...
DLPI errors
service guard requirement violation etc.


So does it means that even you donot want the lan3 part of cluster but in a system its from subnet which is already assigned to lan0 cluster.the system wont work???

[P Arumugavel]

hi,

This is not Cluster behavior only. Even hp-ux will not allow you to keep multiple configured lan cards on same subnet. Routing protocols get confused when two LAN cards in the same system are configured for
the same subnet.

Rgds...

[shabash]

Thanks for the information.


One more question

i want to use 2 Heartbeat IPs.

One local between flex 10 switch.
1 from core back end switch?

whenever i want to configure Heartbeat in between switches,it shows as stationary.

we want it to show as heartbeat.

what should be done to configure single heartbeat IP.

Although i have given it as class A addres for local heartbeat but it always picks it as stationary.

no gateway is defined in heartbeat but still??stationary

any help?

[melvyn burnard]

"Although i have given it as class A addres for local heartbeat but it always picks it as stationary."
Well that has no bearing whatsoever on whether it is a STAIONARY or HEARTBEAT.
All you need do is modify th ecluster ascii file and change any STATIONARY_IP entry to HEARTBEAT_IP, and re-apply the cluster

[shabash]

By renaming the stationary to heartbeat IP will still consider the IP as statitonary IP.

After running cmruncl and cmviewcl -v the primary lan cables are 2 and standby as one.

it means by renaming the statitionary to Heartbeat the stationary cable still send Heartbeat packets???

As per docs.hp.com weblink

whenever an interface configure by service guard cmquerycl command the default scheme will be used.

stationary for stationary data ip not H.B
Heartbeat for data plus H.B

we have tested lan failover and by renaming the stationary ip remain stationary no matter if we rename it.

is there anything missing here???

[Stephen Doud]

Changing STATIONARY_IP to HEARTBEAT_IP in the cluster configuration ASCII file gives Serviceguard permission to send heartbeat messages on the associated NIC configured with the fixed IP.
Configuring multiple/all IP-configured NIC with HEARTBEAT_IP status provides for redundant heartbeat communication paths.

Just remember that if you change the STATIONARY_IP to HEARTBEAT_IP for nodeA:lanX in the cluster configuration file, you must also change it's partner NICs on other nodes in the cluster configuration file.
After making this change, cmapplyconf the file to update the cluster binary file.

Use cmviewconf or 'cmviewcl -v -f line' to inspect the changes.

[Ismail Azad · Accepted solution]

Hi,

Having two NICs on the same subnet for a specific server is one of the checklists in troubleshooting network connectivity. It is highlighted in serviceguard as high availablity revolves around the concept of "redundancy" and in this case NIC cards. Doing this will result in strange behaviour of the cards being able to function independently but_not_simultaneously.

Certain folks come to know about this rule only in a SG environment as the rule "stands out" in SG although it is one of the FAQs in troubleshooting "network connectivity".

> Could any one send link for Hp document regarding this by "highlighting the subnet info line"

> we need to communicated this to customer with proper Hp link.

Read the note on intermittent failures in page 40 of the attached document. That would be enough proof for the costumer as it is the HPUX LAN administrators guide!

> You have assigned points to 12 of 136 responses.

Regards
Ismail Azad

[shabash]

1. It means that whtever interfaces the serviceguard cmquerycl evaluate the network and make certian NICs to stationary_IP or Heartbeat_IP,we can manually edit replacement and service guard wont give any error?

i have run cmquerycl on 3 nodes it will make
lan0=heartbeat(private)
lan1=standy(private)
lan2=stationary
lan3=standby.

#possible standby network for lan0:lan1
#possible standby network for lan2:lan3

usually on other system serviceguard when configured as per requirement give this line as

#possible standby network for bridge network lan0
#possible standby network for bridge network lan1.


2.Now i donot see any bridge comments by service guard but my lans are failover to each other??

Is this ok for service guard lan communication???