Re: Fail Safe Installation Question

Gerold Krommer wrote:
> Hi,
>
> I'm actually surprised about the confusion my post created.
>
> HACMP compares to MS Cluster. Both manage something that I call genericly(?)
> a package.
> A package contains volume groups to be (exclusively) mounted on the active
> node, an IP address to follow wherever the package runs and one ore more
> applications (I know, I'm oversimplifiying but since I got so much from this
> conference I want to donate to)
>
> Fail Safe is something that obviously sits between the MS Cluster Service
> and the package. Why it is exactly needed, I do not know. Unix systems do

It's not exactly needed, it's just something that makes it easier for the proverbial "point'n click" windows admin to set up a working Failover-Group. You can achieve the same by deploying some scripts and setup appropriate listener.ora and such. But why would you want to do that, if Failsafe makes the creation of a cluster group a matter of about 5 Minutes (even if you type very fast, I don't think you can beat that by much ;-) )

> not have anything similar. I have installed such 'hot standyby' solutions on
> HPUX (MC/Service Guard, Solaris (Veritas Cluster) and AIX (HACMP) without
> anything like that. I guess, it is the MS way of doing things differently
> that makes this extra service neccessary.
>
> Why is it cheaper than RAC? Because the license says, that if the instance
> does not run more than 10 days per year on the other node no extra license
> is neccessary.

With RAC in general you have more than one active instance accessing one database. With MS Cluster/FailSafe you only ever have one active instance accessing the database. The '10 days per year' thingy applied to Data Guard if I read the licensing correctly.

>
> In our situation the other node in the cluster runs an Informix application,
> so they back each other up and it is not even an extra HW investment (apart
> from some extra shared disk).

Which is why most people use MS Cluster: the hardware can support both, but in general you want to have dedicated servers. And the increased availability is a bonus. Unfortunately most MS Cluster I've worked with have exactly one shared storage enclosure. And in one case that went down the drain :-(

>
>
> Thanks for the input,
>
> /Gerold
>
Received on Fri Oct 22 2004 - 08:27:02 CDT