-2

I planed to develop a customized HA program for MSSQL, but I got a question when reading some articles about cluster.

When master is down, then slaves vote for a new master, and the new master will take over virtual IP address of the old one.

What is virtual IP address here ?

For example:

A: master 192.168.1.100
B: slave 192.168.1.101
C: slave 192.168.1.102

Is there a another IP address assigned to A? or the IP 192.168.1.100 is the virtual IP ? It is not possible to add extra network cards.

Windows system.

2
  • 1
    Teddy, welcome to SF, but I fear your first question may get closed. At the moment, it looks like a woefully underspecified, completely theoretical question, and those are very hard to answer meaningfully; we much prefer real-world problems with lots of relevant information here. If you can overhaul this question to be more concrete, it might get better answers; you might want to read our guide on asking good questions first.
    – MadHatter
    Dec 31, 2013 at 9:13
  • MadHatter, I am really sorry about that, and I got my answer, thanks for your advise, I will pay attention next time.
    – Teddy
    Jan 2, 2014 at 1:08

1 Answer 1

0

I am not sure exactly what you are up to, but here are some cluster address fundamentals which tend to match mostly, either through being built in or through reliance on external functions to similar effect:

A HA cluster will usually create, in addition to the ip addresses specific to the individual cluster nodes, one or more cluster ip address(es) which belong to the cluster.

Depending on the setup and specific tech this cluster address can commonly either

  • bind to one of the cluster nodes (i.e. be associated with that nodes NIC MAC address), whereby that node becomes the active node. On failover another node will bind its MAC address to the cluster ip address and assume the active role.

  • be associated with a multicast MAC address which all or some cluster nodes can subscribe to, whereby it is possible to have several active nodes. In this scenario the cluster ip address can be a multicast ip address (which makes the binding protocol compliant) or more uncommonly (such as in a Microsoft NLB cluster) an ip unicast address which binds to a multicast MAC address (which requires the network routers/switches to be able to handle this protocol incompliant L3/L2 mapping, usually through extra config directives).

The cluster will commonly use the cluster address only for client-to-cluster traffic, and will use individual node addresses for intra-cluster traffic such as state information and heartbeat. The intra-cluster traffic will sometimes be placed on a dedicated network.

4
  • Thanks, I need to design a "one active node" HA system, and I will listen two ports on each node, one for inner traffic(heartbeat, command, etc), and another to provide service outside. Once master(192.168.1.100) is down, will new master take over IP "192.168.1.100" ?
    – Teddy
    Jan 1, 2014 at 9:54
  • Well that's the general shared cluster virtual ip addres idea, but it is all up to how one configures things. MS SQL comes with quite a few manners and options for building redundancy, here is one way: technet.microsoft.com/en-us/library/hh231721.aspx
    – ErikE
    Jan 1, 2014 at 10:32
  • Thank you, I reviewed the help documents in detail, that is one function in standard edition, so I still need to develop the program, thanks again, that article is very important for me indeed.
    – Teddy
    Jan 2, 2014 at 1:05
  • serverfault.com/help/someone-answers
    – ErikE
    Jan 2, 2014 at 1:57

Not the answer you're looking for? Browse other questions tagged or ask your own question.