Memcached Clusters?

[Steven Grimm]

Moritz Maisel wrote:
> As far as I understand the (perl) client-code, the client just tries to 
> connect to the next server from the pool until it is successful if a connect 
> fails.
>   

That probably is the default behavior; when the Perl client was written 
it was thought to be the right approach. But I recommend you get rid of 
that code from the client. In our experience it causes more problems 
than it solves; we removed similar "try the next server" code from our 
PHP client. Let it fail to the database if a server goes down and you'll 
have far fewer headaches with cache inconsistency.

-Steve