DR Question

[Srinikalyan]

Before failover:
Production IP -> production hostname
Disaster site IP -> Disaster site hostname

After failover i.e production site down:
Disaster site IP -> Production hostname

All the external systems connecting to the production hostname should point to the new disaster site IP. How can we achieve this effectively?
HP-UX 11iv2
Serviceguard and continental cluster

Thanks,
Srini

[Jozef_Novak]

Hello,

hostnames do not change, just relocatible IP addresses of the packages that fail over to another site are moved over to the secondary node in that site. This is done by the ServiceGuard software during the failover. The external systems do not have any knowledge about on which host does the package actually run.

For more info, check the ServiceGuard configuration guide.

J.

[Ganesan R]

Hi Srinikalyan,

Do not confuse the HP-UX host name with cluster package name. In Serviceguard all the resources are combined in a single entitle called package. Virtual IP address and virtual hostname will be assigned to this package when it starts up. End user will connect to this package/server using the virtual ip or virtual hostname. Note serviceguard application users will not use HP-UX host name or ip address.

Now, when the primary node fails, the entire resources bundles as package will be moved to alternate node(disaster node in your case) including it's virtual ip address and virtual host name. So end users will continue to access the applications using the same virtual ip/hostname irrespective of physical servers.

Hope this clear your doubts..

[Srinikalyan]

Let me clear my question:
All the external systems are connected to the production site virtual ip address host name. In case of disaster, external systems should use the same hostname but with different virtual ip address(i.e. in different subnet). I think we can do this with DNS mapping but I don't know how?
Thanks,
Srini

[Steven E. Protter · Accepted solution]

Shalom,

My suggestion would be to change the DNS database.

Do the IP swap there.

SEP

[Srinikalyan]

Thanks Steven,
Can that the IP swap be done automatically in the DNS server during Serviceguard failover?
Thanks,
Srini

[Ganesan R]

Hi Srinikalyan,

Let us clarify few things. We can have multiple virtual ip's/hostname on the physical server where the cluster application is running. You can use any virtual ip/hostname to connect to the physical server, but not to the cluster application.

As far as single cluster application is concern you cannot use different virtual ip address. Means the users connecting to that particular cluster application cannot use different ip's even with the help of DNS. Because in DNS also you can create alias for hostnames, but ip should be same.

Your questions is this..

>>>I think we can do this with DNS mapping but I don't know how? <<<

Yes it is very much possible by creating alias in DNS records. You can create multiple hostname entries pointing to same server. but not with cluster virtual ip's.

[Srinikalyan]

Hi Ganesan,
- Let us clarify few things. We can have multiple virtual ip's/hostname on the physical server where the cluster application is running. You can use any virtual ip/hostname to connect to the physical server, but not to the cluster application.
Does this requires the same subnet in both the sites, isn't it?
In production we have 10.189.39.* subnet and in disaster site we have 10.25.124.*. In this case how can I use the same virtual ip/ hostname in disaster site when production fails.
Regards,
Srini

[Stephen Doud]

Serviceguard patch PHSS_39046 for version A.11.18 implements a long-awaited new feature - cross-subnet clusters. This will enable you to build a cluster of nodes which operates on different subnets. However, when a package is moved to the alternate site and a local subnet-related relocatable IP is awakened, the DNS server must be updated to redirect the "hostname" to the alternate site's relocatable IP.

[Ganesan R]

Hi Srinikalyan,

Are these 10.189.39.* and 10.25.124.* cluster ip's? or Unix ip's?

[Srinikalyan]

Ganesan,
-Are these 10.189.39.* and 10.25.124.* cluster ip's? or Unix ip's?

Both the production node(Unix server) ip and the package ip are in the subnet 10.189.39.0

Both the DR node ip and the package ip are in the 10.25.124.0 subnet.

Regards,
Srini