> On Mar 29, 2016, at 3:20 AM, Pedro David Marco <pedrod_ma...@yahoo.com> wrote: > > it seems not a local performance issue (not virtual and very powerful) but a > remote problem... > > when destination takes too long to send the final OK back, watchdog timeout > happens.. and then > all other smtp deliveries go down... apparently due to no smtp process (as it > exited due to the watchdog) > after some minutes, when gmgr tries again, it works ok... > > My concern is why when one smtp dies (by self-exit or by kill), all > deliveries been processed at that time, fail as well, > when theoretically are performed by different smtp process... may it be that > watchdog timeout kills all smtp at a time?
If you *refuse* to listen, then you're wasting your time here, and that of anyone else not lucky enough to skip your posts. Wildly incorrect theories about how Postfix works will get you nowhere. Watchdog timeouts protect against kernel bugs and DO NOT happen unless your kernel is buggy. If you have not configured ridiculously long smtp timeouts (more than 5 hours), watchdog timeouts can only happen if your kernel fails to honour the requested timeouts on Postfix I/O poll operations. The remote server IS NOT the problem. Further posts along the same lines are pointless, will be ignored, and may ultimately lead to termination of your list subscription. If you want help, post the output of "postconf -n" and system information (operating system and version, virtualization stack details, ...). Perhaps someone has had similar issues on a related system. -- Viktor.