| Subject: | Dead host is not marked as such |
| Date: | Mon, 23 Jun 2008 15:39:05 -0500 |
| To: | bug-Cache-Memcached [...] rt.cpan.org |
| From: | Alex Solovey <a.solovey [...] gmail.com> |
Distribution: Cache-Memcached-1.24
This seems to be reported already:
http://lists.danga.com/pipermail/memcached/2007-July/004796.html
In 1.24, if host is dead when connection attempt is being made,
_dead_sock() does not put it into %host_dead and next request will try
to connect to the dead host again instead of waiting for 20+ seconds
before doing so. It seems that _dead_sock requires %sock_map entry to be
present but that entry will appear only after successful connection.