Warning: in_array() expects parameter 2 to be array, string given in /virtual/theblogs.bestsolution.at/httpd/htdocs/wp-content/plugins/google-one/google-plus-one.php on line 344

Warning: in_array() expects parameter 2 to be array, string given in /virtual/theblogs.bestsolution.at/httpd/htdocs/wp-content/plugins/google-one/google-plus-one.php on line 346

systemd, the Linux dilemma – and why we will migrate away from Debian


Warning: in_array() expects parameter 2 to be array, string given in /virtual/theblogs.bestsolution.at/httpd/htdocs/wp-content/plugins/google-one/google-plus-one.php on line 344

Warning: in_array() expects parameter 2 to be array, string given in /virtual/theblogs.bestsolution.at/httpd/htdocs/wp-content/plugins/google-one/google-plus-one.php on line 346

from https://openclipart.org/detail/203410/misleading-directions

Much has been said about the landing of systemd in most Linux distributions, with Debian and Ubuntu marking another big milestone for systemd rollout.

Aimed being a more or less drop in replacement for the old SysV “init” process, it has caused immense tensions within the Linux community in general.

I wont reiterate all those arguments in favor or against systemd, all of that has been done in countless other blogs, mailing list flame wars, bug reports and many other places. What you can say from a neutral POV is that the discussions around systemd have deeply divided the Linux community. Having been professionally involved with Linux for more than 15 years, I do not remember any other topic causing so much heat and bad energy.

Some time ago I read an article asking why today someone would still refrain from using systemd given the wide adoption is has reached plus the tons of features it offers compared to legacy SysV init systems. It also said that mostly people who had never used systemd were against it, but once you had used it, you would not let it go anymore.

Well, I disagree.

Current situation

Let me first describe our current situation: For the biggest part of our business, we are business software developers, typically based on some kind of 3 tier architecture. Many – if not most – of our customers rely on us to run our software for them on servers that we operate for them. That is not the ordinary “webhosting” business, but instead we offer business critical, highly available services, running mostly on Apache cloudstack these days. Besides some on-premise servers, we have servers in 3 data centers around Europe. A quick look into our OpenNMS installation tells me, that more than 110 servers are currently monitored. The vast majority of those servers run Debian wheezy or LTS squeeze (more than 90%), the rest are mostly CentOS based installations and very few Windows Servers running SQL Server.

The workstation side is more fragmented, with the majority of our developers running Fedora for their daily work, some Debian, some OS/X and even some Windows boxes. I personally have been a Fedora user for a long time, so systemd is no stranger neither for most of our developers nor for myself.

thou shalt not fear the shell

After Debian’s decision to switch to systemd last year, we have been evaluating the situation since. Operating servers can be quite a “delicate” thing and switching core components without very good reasons is something that you avoid most of the time.

“If it ain’t broke, don’t fix it” describes nicely how we handle sysop things.

From our POV, systemd offers very limited beneficial features compared to what we have today. No need for a “journal”, we are perfectly fine with raw (and always accessible) log files. No need for parallel service starts, OS boot times have already improved dramatically over the past years. No need for extra “accountability”, monitoring already works quite nicely. And writing init scripts is easy, if you have done it more than once.

I do understand that systemd may provide useful features for packagers, but for us I fail to see the big benefits required to justify such a major change of a core OS component. Additionally, given how problems about systemd are treated (see [1] or [2] for some recent examples), it really makes me afraid having to rely on a central piece of software that obviously still has major flaws and – equally important – that causes so much hostility.

Alternatives

So, what are our options? Like I said, we have been evaluating alternatives since about January 2015, a long and complicated process. Our main hope is that Devuan [3], a systemd free Debian fork, will gain enough momentum to become a real alternative, both technically and in the long term.

Additionally, we are about to deploy a couple of FreeBSD [4] servers internally in order to get a feeling if that would work as a plan B. This of course would be a major change as well, but FreeBSD certainly gave us the kind of long term stability that we want.

[1] https://bugzilla.redhat.com/show_bug.cgi?id=1213778
[2] https://bugzilla.redhat.com/show_bug.cgi?id=1213781
[3] http://www.devuan.org/
[4] https://www.freebsd.org/

Spread the love

5
Leave a Reply

avatar
4 Comment threads
1 Thread replies
0 Followers
 
Most reacted comment
Hottest comment thread
5 Comment authors
0rbiterNavid ZamaniRobertUdo RaderMarco Carlo Spada Recent comment authors
  Subscribe  
newest oldest most voted
Notify of
Marco Carlo Spada
Guest
Marco Carlo Spada

Very interesting point of view.
Have you considered Slackware? And in this case why is not a choice?
Marco

Udo
Member

Thanks Marco for bringing Slackware into focus 🙂

I have not tried slackware in a long time and must admit that it just slipped through our list of options.

However, at least from what I read about the project’s status quo, slackware definitively deserves a closer (re)inspection and I am in fact doing a slackware “current” installation as I write this reply.

Udo

Robert
Guest
Robert

Since I got ubuntu 14.04 LTS on my Intel nuc, I can boot fast, but shutdown lasts forever. Need to pull the plug. I don’t remember the last time I did this in Windows. So I will definitely leave systemd as well.

Navid Zamani
Guest
Navid Zamani

No need for big discussions. 1. Unix is so successful because its great core ideas. One of those is modularity. Small things, that can be nicely combined, but don’t have to. Another one is simple text files that can be used normally. And a third one is “everything is a file”. 2. If something lacks that, it is not Unix. Monolithic “applications” came with the businessification of software, where the ideal was a monopolistic strangling of the customer, by making it an either-or. (See: Jabber vs. WhatsApp (which is also just a Jabber client, and probably illegally cloned GPL code… Read more »

0rbiter
Guest
0rbiter

One word: Gentoo.

Post Navigation