Error handling in Cache::Memcached

[George Schlossnagle]

Ask Bjørn Hansen wrote:
> 
> On Jul 12, 2006, at 11:53 AM, George Schlossnagle wrote:
> 
> [...]
>> intentional?  Would a patch supporting 'die() on non-recoverable
>> connection errors' be looked on kindly?
> 
> "I rehashed and used another server" would be a recoverable error?  If
> so, then I suspect then in practice for most installations it'll never
> error out.
> 
> Except if you only have one memcached, then I don't really understand
> the use case.  :-)

Intermittent errors?  You have rehashing disabled in the client?
Perhaps you only have one memcached?  Something entirely different that
you didn't expect?

It seems strange to me that this isn't viewed as potentially useful
information to be able to extract from the client.  For instance, in my
case I was writing a little performance test suite to test out what
different access patterns did to performance.  I bumped up against the
default connection limit on my memcached instance, but I didn't know -
even with 'debug' enabled.  It caused me to root around a bit to see why
my numbers went wonky.

This is just one use case, but it seems pretty basic that a client code
for a client-server connection be able to get some basic debugging info
on itself, at least if you set it into a mode where you get that sort of
feedback.

> Brad will be out until the first days in August[1], so he'll likely not
> reply before then.

OK, it's certainly no rush.

George