Combining milter restart with postfix reload sounds like a good suggestion. Our milter restarts due on rare occasions due to upstream changes, but combining it with postfix reload or postfix -> milter seems good solution. Thanks very much for your input.
Thanks, Purush On Thu, May 11, 2017 at 12:22 PM, Viktor Dukhovni < postfix-us...@dukhovni.org> wrote: > > > On May 11, 2017, at 12:16 PM, Purushotham Nayak < > purushotham.m.na...@gmail.com> wrote: > > > > 2. I noticed that when the milter application exits, it sends FIN to > smtpd but smtpd keeps the connection open in the CLOSE_WAIT state. When the > client sends something new and smptd tries to send SMFIC_ABORT to the > milter on the old connection in the CLOSE_WAIT state it gets a reset and > then the connection is closed. This is not really a problem, but just that > the socket stays in CLOSE_WAIT state sometimes for a long time if there is > nothing new from the client. > > The smtpd(8) process is single-threaded, and only looks at the milter > connection > when it is using the milter. If the milter connection closes while > smtpd(8) is > doing other things, it will be kept open until it is time to try to use it. > > If you are in the habit of restarting milters, you may as well at least > "postfix > reload" when you do so. A more invasive: > > postfix stop; delay?; <milter> stop; <milter> start; delay?; > postfix start > > will ensure no anomalies in milter use while the milter is down. > > -- > Viktor. > >
