Hi Pieter, On Mon, Nov 20, 2017 at 10:24:04PM +0100, Willy Tarreau wrote: > On Mon, Nov 20, 2017 at 10:18:31PM +0100, PiBa-NL wrote: > > Hi Willy, > > > > Op 20-11-2017 om 22:08 schreef Willy Tarreau: > > > OK thank you. I suspect something wrong happens, such as the master > > > killing the same kevent_fd as the other ones are using or something > > > like this. > > > > > > Could you please try the attached patch just in case it fixes anything ? > > > I have not testedit and it may even break epoll, but one thing at a > > > time :-) > > > > > > Thanks, > > > Willy > > > > You patch fixes the issue. > > Great, thanks. > > > If you've got a definitive patch lemme know. > > OK, I'll check with William because I don't know the master-worker code > yet and don't want to start to put my fingers everywhere :-)
We found that it's the first fork_poller() placed after the fork which fixed the issue but that makes absolutely no sense since the next thing done is exit(), which also results in closing the kqueue fd! So unless there's something in at_exit() having an effect there, it really doesn't make much sense and could imply something more subtle. Thus my question is : do you have a real use case of master-worker+daemon on freebsd or can we release like this and try to fix it after the release ? It's about the last rock in the shoe we have. Thanks, Willy
