Hello List,

We're seeing a small but steady number of complaints about problems connecting to authdaemond (example follows):

Jan 9 18:15:11 XXXX pop3d: [ID 702911 mail.error] authdaemon: s_connect() failed: Connection refused Jan 9 18:15:11 XXX pop3d: [ID 702911 mail.error] [Hint: perhaps authdaemond is not running?]

We're running authdaemond with a Postgres database backend which we're pretty sure is not overloaded and have already increased the number of authdaemonds to 20, with no noticeable change. The server running courier-imap/courier-authlib has plenty of spare CPU capacity.

Could this be an instance of the following problem alluded to in the default authdaemonrc?

=> # PLEASE NOTE: Some platforms may experience a problem if there's more than => # one daemon. Specifically, SystemV derived platforms that use TLI with => # socket emulation. I'm suspicious of TLI's ability to handle multiple => # processes accepting connections on the same filesystem domain socket.

Solaris is allegedly SystemV based after all?

Or is something else going here?

Any comments would be very much appreciated....

Thomas