systemd: bugreports for missing service-files

[Adam Williamson]

On Tue, 2011-06-21 at 01:18 +0200, Reindl Harald wrote:

> what exactly do you not understand in the fact that even starting myqld
> first does not help with a large database since mysqld needs some seconds
> to accept connections and systemd fires up the follwing services

So would any other init system; if the startup script has returned
success, any init system figures it's now fine to move on to the next
one. How else would you design it, exactly?

I can see a scenario where in a sequential init system the daemon
happens to get into its 'working' state by the time any other service
that needs it is fired up because the other services that need it happen
to be way down the stack, but that's _pure dumb luck_ and can't be
relied upon at all; what if you streamline your service set a bit and
now the dependent services fire up 'too early'? What if you get a faster
disk? The fundamental problem in such a situation would be the startup
script for the daemon saying it's ready to accept connections - i.e.,
saying it's started up - before it actually is. That would be the
appropriate bug to fix.

I still can't tell from your ranting exactly what it is you feel systemd
does wrong, and from his replies, neither can Lennart. What exactly is
it that you think systemd is doing which upstart didn't?

> so where was the QA before the release which forbids fix the release?

Well, a few things:

There's no planned testing of mysql prior to release. QA does not have
the resources to test every feature of Fedora; what gets tested are
things that are critical to the basic functionality of the system, which
does not include mysql. Anything which can be fixed with a post-update
release is far less significant than things which can't. If you want to
see what things we actually undertake to have working in Fedora
releases, see https://fedoraproject.org/wiki/Fedora_Release_Criteria .

You are misunderstanding this whole 'forbids fix the release' thing.
What is forbidden is converting a service from sysv to systemd-native
within a stable release, because that's a major change which should
happen between releases. It should never be necessary to do such a
conversion to fix a bug, however. Any bug should be fixable by improving
the sysv service script and/or fixing any relevant bugs in systemd;
there should be no situation where behaviour regresses between F14 and
F15 in such a way that it can't be fixed in any way except by migrating
a service to systemd-native.
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Fedora Talk: adamwill AT fedoraproject DOT org
http://www.happyassassin.net