Boguk Maxim wrote:
Somehow first mail didnt get to maillist. So second attempt:
After update Perl memcached API to latest version i got some troubles
with disconnects to memcached server:
Both of these function trying clean $sock_map{\$sock} !
So... Connect code set $sock_map{$sock} and cleanup code trying clean
$sock_map{\$sock} !?
And of course dead sock never cleaning at all and API try use dead sock
again and again.... (because bad socket permanently cached in %sock_map)
This is really painful issue leading instability of projects.
Yes, I came to the same conclusion when adopting memcached. See my bug report
in the CPAN RT system[1]. The fix indeed is to remove the backslashes.
By the way, I'm still suffering from the occasional segfault in long-running
programs, when one memcached instance goes offline. Must be something in the
socket handling, but I haven't had time yet to investigate.
Rhesa
[1] http://rt.cpan.org/Public/Bug/Display.html?id=27181
