is samba LinuxVirtualServer-able?

John Rodkey rodkey at westmont.edu
Fri Oct 26 11:29:02 GMT 2001


On Fri, 26 Oct 2001, Christopher R. Hertel wrote:

> >  
> > I'm still interested in finding out if the client can reconnect again
> > seamlessly to another server after client or server failure.
> 
> There's nothing in the architecture that makes this easy.  The client is
> going to make a TCP connection and, if that connection is lost, the client
> will barf rodents just as you'd expect.  I suppose you could create a
> stateful proxy, but that still means a single point of failure and a
> single bottleneck.

I don't quite understand the TCP connection being lost problem.  I have
a number of SMB clients that connect to a server, and if the server goes
down and comes back up, their 'connections' are still apparently valid.
Now, I assume they must have been renegotiated when the server realizes
the client expected a file to be open, but the user never realizes his
smb server has gone down and been replaced with another.

> Even if another server could 'spoof' the connection, SMB is a stateful 
> protocol and the second server would need to replicate the server-side of 
> that state.

In my experience, the problem really comes when you're writing a file or
it's locked for writing, anyway.  Presumably all the state information
associated with writing a file (locks, etc) are only kept in memory
and not on a disk, which could theoretically be shared among many realservers.

> Chris -)-----
> 
> 

-- 
        John Rodkey, Information Technology, Westmont College
			rodkey at westmont.edu





More information about the samba-technical mailing list