[Catalyst] Sessions and memcached storage

[J.G.Konrad]

I've been having problems with session storage using memcached with more
than one memcached instance.  The sessions are expiring before they are
supposed to.  If there is only  one memcached instance everything works like
it is supposed to.

Has anyone else had this problem?

One of my colleagues did some stress testing and he came with some evidence
that the internal cache ( buck2sock ) is incorrectly mapping to sockets
under stress. That is to say the value out of the buck2sock cache is
different than the value from sock_to_host() which should not be the case.
That could help explain the problems with multiple memcached instances but
our lost session also happen when the system is not stressed.

Thanks,
  jason
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.scsys.co.uk/pipermail/catalyst/attachments/20071204/a0f92=
a46/attachment.htm