Dear William-san, Sorry for the late reply.
> I'm not convince that it should be done within HAProxy, you will still > have the same problem if the master crashes. I understand that it is the same when the master crashes. The remaining pid file is unavoidable, but I think it is necessary to check if the PID written in the pidfile is the haproxy's PID. Otherwise, the -sf/-st option could kill other processes. I will send a patch to solve this problem. Best regards, MIZUTA Takeshi > -----Original Message----- > From: William Lallemand <[email protected]> > Sent: Tuesday, October 13, 2020 5:42 PM > To: Mizuta, Takeshi <[email protected]> > Cc: 'HAProxy' <[email protected]> > Subject: Re: [PATCH] BUG/MINOR: mworker: delete the pidfile when the > master process is stopped > > On Thu, Oct 08, 2020 at 04:59:46AM +0000, [email protected] > wrote: > > Hi, all, > > > > haproxy does not delete the pidfile when stopped. > > If the PID in the remaining pidfile is reused by other process, > > the operation using the pidfile may affect other process. > > In master-worker mode, fixed to delete pidfile when stopped. > > > > However, in the daemon mode, it has not been dealt with yet. > > Is there any objection to dealing only with master-worker mode? > > > > Best regards, > > MIZUTA Takeshi > > > Hello, > > I'm not convince that it should be done within HAProxy, you will still > have the same problem if the master crashes. > > It's also a big change of behavior that could break existing scripts. > > In my opinion this should be done this in your init script. > > -- > William Lallemand
