On 18/06/2016 23:20, Guillermo wrote:
> And for release 0.21, the authors have also rolled their own
> experimental process supervision tool: the supervise-daemon(8)
> program.
> (...)
> start-stop-daemon(8). Specifying a PID file (by assignment to
> 'pidfile') is mandatory, probably so that the openrc and rc-service
> programs know which supervise-daemon process to kill when the
> corresponding service is wanted down.
And once again, a service manager chooses to develop the same dirty,
unreliable hacks as all the mainstream solutions have done in the past
few decades and which have led to the rise of alternatives, *including*
OpenRC.
All that wasted effort, which could have gone into integrating OpenRC with
a real supervision suite. But no, it's so much better to reinvent the
square wheel.
> Yeah, I can hear the screams
> :-D, but I guess OpenRC can't do it any other way without a major
> redesign.
It could.
It could require a supervision tree to be operational before running, and
integrate with it the same way s6-rc does with s6: by translating start and
stop calls into commands sent to the supervisor. If they had bothered to
contact me, for instance, I would gladly have offered assistance.
But I guess it's easier to cook your own half-assed hack than, you know,
communicate with a human being. Brrrr, the horror.
> Anyway, I thought it still deserved a mention :)
It does deserve a mention: "never use it".
I may sound harsh, but I'm f*cking sick and tired of seeing the same
mistakes pop up again and again in order to tackle what is a *solved problem*
and has been so since daemontools in 1998.
--
Laurent
Received on Sat Jun 18 2016 - 21:41:55 UTC