[nSLUG] XEN & Heartbeat

Ian Campbell ian at slu.ms
Wed Jun 3 15:36:27 ADT 2009


On Wed, Jun 03, 2009 at 06:32:19PM +0000, Ian Campbell wrote:
> On Wed, Jun 03, 2009 at 02:26:37PM -0400, Stephen Gregory wrote:
> > On Wed, Jun 03, 2009 at 06:10:35PM +0000, Ian Campbell wrote:
> > > 
> > > Yeah, but those aren't "fault tolerant" to the point where if you're
> > > writing to disk on one machine which crashes halfway through it
> > > continues seamlessly on the other, are they?
> > 
> > In theory the setup will survive. It is desinged to survive bad
> > hardware. It does not to protect against bad software. If on vm
> > crashes then the other will crash as both are identical and running
> > the same code.
> 
> Yeah, I'm not talking about just maintaining service. Oliver was
> mentioning that a downside of his approach was that existing
> connections would die. There would also be half-finished writes. The
> service would still be 'up', you'd just have to reconnect etc.
> 
> My question was if there was anything out there that DIDN'T have that
> flaw... ie I'm writing 10GB to disk, side A fails, side B completes
> the loss with no interruption from my point of view.

Oops. Side B completes the WRITE with no loss and no interruption from
my point of view.



More information about the nSLUG mailing list