Are there and Built-in Application Failover Technologies in .NET and/or Windows Server 2008 (or SQL

[Charles]

Hi all

A bit of a general question, hence the cross-posting.

I have a .NET application running on a Windows Server 2008 server. For
resilience, I want to run it on another server simultaneously, but only one
instance of the application can be 'active' at any one time. In the event
that the active instance of the application fails for some reason, I want
the passive one to take over. What techniques are there available to make
this an automatic process?

Of course, I can roll my own, as it were, and I would be interested in how
other people would implement a bespoke solution, but are there any built-in
mechanisms in .NET or Windows Server, or indeed SQL Server, that will make
this more straightforward?

TIA

Charles

 

[John Bell]

Hi all

A bit of a general question, hence the cross-posting.

I have a .NET application running on a Windows Server 2008 server. For
resilience, I want to run it on another server simultaneously, but only one
instance of the application can be 'active' at any one time. In the event
that the active instance of the application fails for some reason, I want
the passive one to take over. What techniques are there available to make
this an automatic process?

Of course, I can roll my own, as it were, and I would be interested in how
other people would implement a bespoke solution, but are there any built-in
mechanisms in .NET or Windows Server, or indeed SQL Server, that will make
this more straightforward?

TIA

Charles

This is sounds like you need to look into clustering
http://msdn.microsoft.com/en-us/library/ff650328.aspx#

John

 

[Charles]

Hi John

Thanks for the reply. I have looked at the link and it appears to be a
discussion on the generics of failover clustering, which is really what I
was meaning by "roll my own".

The token idea is interesting, though, as it would seem that if neither
server can establish the owner of the token then they would both assume the
role of master; for example, where the servers haven't crashed but the
communication between them is lost.

Charles

 

[John Bell]

Hi John

Thanks for the reply. I have looked at the link and it appears to be a
discussion on the generics of failover clustering, which is really what I
was meaning by "roll my own".

The token idea is interesting, though, as it would seem that if neither
server can establish the owner of the token then they would both assume the
role of master; for example, where the servers haven't crashed but the
communication between them is lost.

Charles

I don't think they could both be live as the resource could only be
assigned to one node. But it is possible to get into a state where the
are constantly swapping over, which is why you need good comms between
the nodes.

John

 

[Charles]

Hi John

In my particular application, I'm thinking of a socket to a [third-party]
remote server. My application would attach to the remote server, via TCP,
and receive data. The remote server is designed such that many clients can
attach and receive [the same] data, so at my end I only want to attach one
client at a time. I have to decide, for my self, which client app. That's
why I want to reproduce some sort of failover application, ideally using
existing technology, but otherwise using a standard, well-proven technique..

Charles