Re: [PLUG] LINUX ADMIN ADVICE

[Steve Litt <slitt@troubleshooters.com>]

On Sat, 4 Aug 2018 05:27:30 -0400
Rich Kulawiec <rsk@gsp.org> wrote:

> On Thu, Aug 02, 2018 at 07:23:07AM -0400, Rich Freeman wrote:
> > Sure, but while you're busying fixing bugs it doesn't hurt to have
> > the service still running, vs paying somebody to babysit it, or
> > writing custom scripts to babysit stuff for you.  
> 
> Restarting a service which has provided a demonstration proof that
> it's broken isn't a good operational practice.  It *does* hurt to
> have it still running.  It should be shut down until such time as the
> problem is reproduced, isolated, understood, debugged, and fixed.
> 
> > Proving that your software doesn't have bugs is a near-impossible
> > task.  Defense in depth makes a lot more sense.  And as I mentioned
> > systemd does a lot more than just restart services.  
> 
> 1. Blindly restarting a failed service isn't a "defense" of any kind.
> It's wishful thinking.
> 
> 2. I'm well aware that systemd "does a lot more".  But that's not
> a feature.  That's a catastrophic design blunder, one which is
> obvious to anyone with even a rudimentary grasp of the "Software
> Tools" and Unix philosophies.
> 
> 3. You do know that this isn't my first day on the job, yes?
> 
> > IMO if your requirements are so simple that rc.local does everything
> > you need, there is a good chance that rc.local is already too
> > complex for you.  Why not just make init a bash script?  
> 
> I think you're missing the point.  If your requirements, for a value
> of "you" meaning "pretty much all of you", aren't that simple or
> nearly so, then you're probably doing it wrong.  The solution to that
> isn't to layer on yet more complexity, it's to step back, re-evaluate
> the design, and consider how it could be done more elegantly, simply,
> and robustly.
> 
> This requires more work.  It may require throwing things away that had
> substantial effort invested in them.  It sometimes requires going
> back to a blank whiteboard.  It may be tedious and expensive.  But
> it's what I'd expect competent professionals to do.

If you want a crash to stay crashed, you can set it to do so in systemd
or s6. You can do it in runit with a simple forever loop hack or a
downfile hack. OpenRC has no facility (in and of itself) to restart. So
the person wanting single-shot behavior will always be able to get his
way, regardless of init system.

SteveT
 
Steve Litt
Author: The Key to Everyday Excellence
http://www.troubleshooters.com/key
Twitter: http://www.twitter.com/stevelitt

___________________________________________________________________________
Philadelphia Linux Users Group         --        http://www.phillylinux.org
Announcements - http://lists.phillylinux.org/mailman/listinfo/plug-announce
General Discussion  --   http://lists.phillylinux.org/mailman/listinfo/plug