Urgent Help on VCS cluster

[RBK001]

Hi All We have a deployment using SFHA. Mentioned by our partner, the cluster services will be failover if (1) all the NICs are down, and (2) if the healthcheck on a reference Ip is failed. We would like to add a secondary reference IP in the setup, how I can add this information? Appreciated if there is an expert to share the step on this. - rbk

[Marianne]

I'm not sure if the concept of failover was explained properly...
There are a number of things that will cause the cluster to fail over. Any critical cluster resource that fails will cause a failure of the Service Group and failover to another node in the cluster. A set of resources that makes up an application is grouped in a Service Group. These resources are tipically a NIC, a virtual IP, disk storage, mount points and the application that needs all of these resources.
The service group will also fail over if a system failure is detected. A system failure is assumed when communication on both cluster interconnects (heartbeats) are lost simultaneously. It is therefore crucial that the interconnects are connected via totally independend NIC's and completely separate network infrastructure. VCS communicates on the interconnects using its own propriatary protocol and no IP addresses are configured on these NIC's.
The 3rd NIC in each cluster node are used for connection to the public network, each with its own fixed IP address. Each service group that will be configured in the cluster will have its own virtual IP address that is linked to this 3rd NIC.
With the above in mind - do you want to add another IP address in a service group or add another service group with it's own IP?

You did not mention your O/S and SH/HA version? I'd like to provide links to documentation, or else you can start here:
https://vos.symantec.com/documents

[RBK001]

Thanks for the information. We are using Solaris 10 and SFHA version should be 5.0 MP2.

I understood from the vendor that there is a reference IP that will also cause the cluster services to be switchover if it is not accessible after the health check. However, I am not able to point it ou from the configuration. And I would like to add a secondary reference IP for this purpose.

any suggestion?

[Marianne]

Please ask the vendor to explain 'reference IP'. I too work for a Symantec reseller and regurally assist with customer installations. I have no idea what he calls 'reference IP'.  I'm not sure if this is his way of explaining the cluster interconnects/ Private network.  Please ask for explanation of 'health check' as well. Not sure if he is referring to the regular monitoring of resources or the constant 'heartbeats' on the interconnects...

 Please consider installation of 5.0 MP3 as a minimum. SF/HA 5.1 was released about six months ago - my recommendation is to install this version. (We have a number of customers using this version in production; it has lots of improvements that I provide links to.)

VCS requirements are explained in the VCS installation Guide. Link to 5.0 MP3 install guide: http://sfdoccentral.symantec.com/sf/5.0MP3/solaris/pdf/vcs_install.pdf
"Setting up the private network" is covered on p.36.

Cluster failover behaviour is explained in the Users Guide: http://sfdoccentral.symantec.com/sf/5.0MP3/solaris/pdf/vcs_users.pdf. See Section III : VCS communication and operations.
p. 351: About inter-system cluster communications. Heartbeat explanation starts at the bottom of p. 352.
p. 370: About cluster membership and data protection without I/O fencing
p. 382: About VCS behavior on resource faults

[Marianne]

One more thing - Best if we move this thread to the VCS forum...

[Gaurav_S]

Just presuming, I think reference IP relates to base IP in IPMP ?

Can you confirm, what agent you are configuring ? MultiNICA/IPMultiNICA  OR MultiNICB / IPMultiNICB  ?

I believe it should be later one.... however can you confirm if you are using solaris IPMP ? If yes, what is the config for that ? paste below outputs from all nodes in cluster

# ifconfig -a  
# cat  /etc/hostname.xxx     (for interface that is part of IPMP)


Gaurav

[Roger_Zimmerman]

Hi, RBK001, is it possible that you talk about an IP adress somewhere out on the network when you say "reference IP", and if this one is not available, the cluster is switching the service group to another node?

Then you can find this "reference IP" in the MultiNICB, MultiNICA or NIC ressource (depending on what you have on duty) in the attribute NetworkHosts. And this attribute can have more than one entry (its a string-vector), so you can provide more than one "references".

But take care for not using too much of those external IP's, there's a lot of pings out there. And also do not use some "easy-switch-off-devices", because if they are not available the cluster is switching, do you remember ....

Just an idea...
Roger