Hi Wietse, On Thu, Sep 06, 2012 at 02:33:55PM -0400, Wietse Venema wrote: > Naval saini: > > Sep 6 16:37:48 r22 postfix/postscreen[3992]: warning: haproxy read: time > > limit exceeded > > Sep 6 16:46:51 r22 postfix/postscreen[5233]: warning: haproxy read: time > > limit exceeded > > Sep 6 16:50:49 r22 postfix/postscreen[5448]: warning: haproxy read: time > > limit exceeded > > Sep 6 17:02:18 r22 postfix/postscreen[5791]: warning: haproxy read: time > > limit exceeded > > Your load balancer does not send connection information within the > time limit (the default, "postscreen_upstream_proxy_timeout = 5s" > should be sufficient). > > Perhaps the load balancer uses a different protocol than HAPROXY. > > You can use "tcpdump -s 0 -w /file/name" to record a failed session > and see what protocol the load balancer actually uses.
That's something totally possible. Naval seems to be using latest haproxy development snapshot. I think it would help if problem reports on the postfix ML were done with "more reliable" haproxy code than latest devel snapshot, and vice-versa. My non-reg tests have shown that the PROXY protocol in this version seems to work for me but this does not mean there is no bug. A tcpdump trace is clearly needed. Naval, if your trace shows there is no PROXY protocol in the connection, please try again with 1.5-dev11 and bring the issue back on the haproxy ML. Regards, Willy
