[nSLUG] How to avoid systemd in Debian 8 "Jessie"

D G Teed donald.teed at gmail.com
Tue Apr 28 14:44:30 ADT 2015

I have zero experience with systemd.  I have purposely avoided installing
Redhat/CentOS 7 due to preferring to wait and see.  It isn't just
the distro packages, but third party and home made services
I have to be concerned with.

I do have experience with Solaris Service Management Facility, started
with Solaris 10, and I anticipate systemd will work in similar way.

What we lose with systemd is the ability to easily tweak parameters within
the init scripts.  There are many times I've found some check in
an init script I want to skip because it doesn't pertain to a special
situation (e.g. PID file check, base dir, etc.).  With an init script
I am in the driver's seat in seconds.  Without a script, I'm left to google.

Given the choice of direct intervention and getting a service in action quickly,
or searching the net and hoping to find good information that isn't from 2003,
I'll take the intervention option every time.  Say you have a binary that is
running and you want to know the service that starts it.  On init
system I can grep in /etc/init.d for the binary or related bits.  What
are my options in systemd?  I'm sure there are options, but they
are not known.  We system administrators are always in novel
situations where something goes wrong in a unique way.  We need
a tool belt where generic knowledge can be used to investigate.
Plain readable shell scripts was a big chunk of that generic
capability allowing us to fix stuff we had never seen before. or allowing
capabilities to add tracing, etc.  So it makes sense why people
can be upset with the landscape being tilted.

The other issue I have with systemd are the claims.  Faster start up for
example.  Without systemd, my Debian 7 systems are able to reboot on
a KVM or VMware instance in 14 seconds.  On physical hardware,
the big wait isn't for your OS - it is for your BIOS, your
controllers, ILOM, etc.
I fail to see how shaving a few seconds off a 14 second reboot cycle
is worth the switch to systemd.  On top of that, when do you need to
reboot?  With the 4.0 kernel under development, there will be updates
to the kernel without a reboot.  So systemd solves a problem with
boot up time that didn't exist.

A good evaluation of systemd will come on the day you have to
download a random tarball from sourceforge, or Github, etc., and set up
a new systemd service file, after finding none on the net.  Then tell us
how you liked working with systemd.

It's not unlike other changes in IT.  What do you prefer, /etc/shadow
or building
and maintaining an LDAP system?  Or php+mysql, versus node+mongodb?

More information about the nSLUG mailing list