Proposed generic release criterion: service manipulation

[Adam Williamson]

On Sun, 2014-07-20 at 12:20 -0400, David A. De Graaf wrote:
> On Fri, Jul 18, 2014 at 02:43:41PM -0700, Adam Williamson wrote:
> > On Fri, 2014-07-18 at 15:01 -0400, David A. De Graaf wrote:
> > > On Fri, Jul 04, 2014 at 12:22:40PM -0700, Adam Williamson wrote:
> > > > 
> > > > System service manipulation
> > > > 
> > > > It must be possible to start, stop, enable and disable system services
> > > > using the initialization framework's standard commands.
> > > 
> > > Oh yes, that's perfect.
> > > 
> > > Had this been applied to the most basic Linux command - 'reboot' -
> > > we might never have seen Fedora 18, 19, or 20.   :-)
> > 
> > Er, how do you mean?
> > 
> > We have a criterion covering reboot functionality already:
> > 
> > https://fedoraproject.org/wiki/Fedora_20_Beta_Release_Criteria#Desktop_shutdown.2C_reboot.2C_logout
> > 
> > though I note there's a funny inconsistency - at Alpha we require
> > console-based shutdown to work, at Beta we require graphical shutdown,
> > logout and reboot to work, but we never explicitly require console
> > logout and reboot to work.
> > 
> > So, let me propose that we amend the beta criterion:
> > 
> > Shutdown, reboot and logout
> > 
> > Shutting down, logging out and rebooting must work using standard
> > console commands and the mechanisms offered (if any) by all
> > release-blocking desktops.
> > 
> > Thoughts?
> 
> Seems more inclusive.  I don't understand why the title was
> "DESKTOP shutdown, reboot, logout" anyway.
> Should have been just "System shutdown, reboot, logout", or better,
> just omit the first word modifier altogether, as you propose, Adam.

The reason was that it really did only cover desktops; I think I thought
the console commands were covered at Alpha, but as I noticed when you
brought it up, only shutdown actually is.

> My original carp was due to the major aggravation introduced by
> systemd - the inability to cleanly 'shutdown' or 'reboot'.
> Sometimes these incidents manifested as an interminable delay of
> many minutes; other times a complete freeze.  

> This made the "Magic SysRequest" command set essential - which should
> NEVER be needed.
> I believe these failures can all be traced to NFS mounts that could not
> be cleanly umounted.  But what can be more stupid than waiting for a
> non-existent NFS server to respond?  Cleanliness is not an option.
> 
> I propose that the shutdown++ criterion include prompt
> closeout of NFS file systems, requiring perhaps a default
> configuration of autofs mounts that will allow quick umounts, despite
> unexpected inaccessibility of a remote machine, if this is even
> possible.  It always worked before systemd.

I think that's going a bit beyond release criterion material personally,
but let's see what others think. Have you filed this as a bug? It would
help to have full details of how the NFS mounts are configured.
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net
http://www.happyassassin.net