:: Re: [Dng] Dng Digest, Vol 5, Issue …
Forside
Slet denne besked
Besvar denne besked
Skribent: Martijn Dekkers
Dato:  
Til: dng
Emne: Re: [Dng] Dng Digest, Vol 5, Issue 11
Hi,

The reason why RedHat develops and enforces systemd is clear to me: they
> have customers paying for a ready-made system that RedHat would maintain.
> They want to increase their productivity by introducing tools which
> automate things as much as possible, plus security-related features -- a
> valid sales argument. It makes full sense; that's their busyness model.
>


I (my company) looks after a lot of servers, and we are *always* looking
for ways to make things go better, faster, and stronger :) automation is a
*huge* part of what we do, as is security. Systemd does not really offer us
much that we don't already have, and the "price" we need to pay to take on
systemd is not one we are comfortable with. Untested, unproven, and at the
end of the day an overall framework architecture that we think sucks.
Systemd offers a lot of interesting things for desktop users, which we are
not. We are server monkeys, and don't really care about desktop bootimes. I
*like* initscripts, but also like OpenRC and upstart (we are an Ubuntu
shop).

We use a massive collection of tools, as do most people on this list, on a
daily basis. Saltstack is a major piece of our management "backbone" as are
various flavors of hypervisors and containers. Python is the glue we use to
keep it together, as are bash scripts, sed, awk, etc. etc.

I have personally been an active Open Source user, advocate and contributor
for a very long time. I have no problems, however, with using proprietary
software if the need is there - we have a business to run, and we must be
pragmatic.

The point is that we will *always* strive to use the *best* tool for the
job. I have no time to get sentimental or religious about these things. If
a new tool comes around tomorrow that will deliver tangible, measurable
benefits and improvements for our technology stack, I will not hesitate to
throw out whatever we need to in order to take advantage of it, provided it
is good for the business.

Systemd doesn't meet any of our needs better over the existing technology
stack we currently have, and has a lot of fundamental architectural and
organisational weaknesses that make it unsuitable for us to deploy. I have
no problem with people wanting to use systemd - if it works for you, than
more power to you. I do have a very, very serious problem in having this
technology shoved down my throat, with very few options. I do have a
problem with the fact that our team now has to divert very precious and
expensive resources to investigate BSD, and see what we are going to do to
fill the technology gaps between what we have now, and what BSD lacks
(OCFS2, for example).

Personally, I have no emotional bone to pick in the whole Debian systemd
debacle. However, when Shuttleworth made the announcement that "Ubuntu
follows Debian, if they go systemd then so do we" the rug was pulled out
from under us. I love the Linux kernel, for all its issues and
imperfections. From all the "no systemd here" efforts and distro's, Devuan
is the most promising. None of the others (slack, *too, and alpine)
currently meet *our* needs, and we do not have the engineering capacity at
the moment to get them to a point where we would be able to realistically
get to point that is good enough for us to work from.

I am hoping to see a Devuan release soon, so we can gauge the work and
effort required to "ubuntufi" Devuan. If that is likewise too much work
then we will have no choice but to bite the bullet and leave this whole
clusterfuck far behind us, and go BSD. All this bullshit with systemd
pushed us at least 6 months back in our planning and deployment, and cost
us a lot of money. At least we will know for sure that it will a good while
yet before the insanity spreads to BSD. And for that, I do not have a warm
place in my heart for the guys at RedHat.

If you think that this kind of radical, rapid change goes unnoticed in the
enterprise, or that the enterprise doesn't mind such radical changes in
such a short period of time, think again. Few of RedHat's (or Canonicals',
for that matter) clients are impressed with how quickly, aggressively, and
pervasively systemd was pushed down our throats.