Can floating IP address be used out of the Package?

[komodo_1]

Hi, all
I have a problem with Novel eDirectory on HP-UX.
I need to provide HA LDAP service for our appliction. Cause eDir is non clusteraware application, I cannot use MC/SG to cluster the LDAP server. The best eDir HA solution is to make two LDAP server replicated form each other. However, the problem is that I have to make every application able to failover and failback between two LDAP Server IP address. It is impossible for some 3rd party application.
Novel's suggestion is to make there LDAP server bind to floating IP address.
I am not sure whether it work. I can setup dummy package to make a floating IP address. However, I am not sure whether the application on the two nodes can successfully bind to the floating IP address when they start.
I need your suggestion on whether it works and the internal mechanism behind it if it works.

[D Block 2]

komod, you say, "I cannot use MC/SG to cluster the LDAP server."

my response is, "can you use a Single-LDAP Server" given a Non-MC/SG arrangement ?


Also, can you clarify, what do you mean buy saying, "can you make the two LDAP server replicated form each other" ??

My point is, given LDAP, what is your plans ? Do you want the MC/SG nodes to "LDAP Replicate Each Other ?"

And what do you use on each HP-UX server, ?

Sorry, but can you provide more details ?

[Thomas J. Harrold]

Why not create a basice MC/SG package that does nothing more than failover an IP address? If both SG servers are running LDAP, the virtual IP will always be pointing to a valid LDAP server. Use the virtual IP in your LDAP dependent applications.

If you have MC/SG installed, this would take only a short time to setup...

-tjh

[komodo_1]

Thanks Thomas. Your suggestion is what Novel said. However, I have no hardware available now, cause we are still in architeture desgin phase. I just wondering the mechanism behind the soluiotn.
As you said, float IP is available only in one node. How did two LDAP server successfully bind to the float IP. Say, If float IP is on node A, than the LDAP server on node A will succeed in socket operation, that make sense. However, why LDAP server on node B will suceed in binding to the float IP? And what is the case, when float IP switch to node B?
Thank you in advance

[Christopher Caldwell]

Floating IPs aren't intrisic to ServiceGuard; floating IPs are simply a technique you can use to bind more than one IP address to an interface.

You can use ifconfig to add additional IP addresses (there are caveats, like all IPs have to appear in the same IP network).

If the application is written to bind to specific IPs then it will work.

Other applications use the INADDR_ANY constant to bind to all IPs on an interface. In that case, the app listens on all interfaces, so it will listen on the "floating IP" but it will speak from the base IP.

ServiceGuard buys you some handy tools to allow IPs to migrate from host to host or NIC to NIC given certain failures.

You might consider a load balancer in front of a number of LDAP servers for availability. Allow the LDAP servers to replicate behind the scenes.

[komodo_1]

thank you for you answer. I am now clear about what I shall do