Le 02/11/2015 03:31, Go Linux a écrit :
I thought some of you might be interested in this post about the article posted
below:
http://forums.debian.net/viewtopic.php?p=597967#p597967
I couldn't begin to understand the original treatise but could kind of follow
what tomazzi was saying. Any thoughts?
golinux
I disagree with the view of the author about starting services. I
quote below what he expects from a service supervisor:
Tomazzi:
<< When the service is starting to write to a socket, it means that it's
ready: configuration was read and initialization is done: Stop the
service (SIGSTOP). Wait for other services to start. Continue the
service process (SIGCONT). That's all.
The SIGCONT is sent to all services once all of them are started.>>
Servers are able to deal with the client not being ready
(obviously); and the clients should also be ready to wait until the
server is ready. The only thing init should ensure before starting
services is that the necessary mounts are done, and this has no relation
with socket activation.
Therefore, all this checkpoint machinery is crap. I think this is
part of the crap systemd fans have endeavoured to broadcast all around,
and even some anti-systemd people haven't wiped it out of their mind
yet, because it's not so easy.
Reporting readyness is admin-friendly, but this can be done
trivially, in the s6 fashion; it does not take a library to do.
Didier
_______________________________________________
Dng mailing list
Dng@lists.dyne.org
https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng
