My two cents on "the new Linux" -> RedHat EL7 / CentOS 7.

A 5 star hangout for overworked and underpaid system admins.
sblantipodi
Posts: 252
Joined: 2009/07/10 09:43:13
Contact:

My two cents on "the new Linux" -> RedHat EL7 / CentOS 7.

Post by sblantipodi » 2014/07/18 18:04:06

I start this thread because CentOS 7 leaved a strange disappointment on me and because I am going to stay on the train for 6 hours.
7 is a huge step forward over previous releases and I was ready for big changes.

Let's start to summarize only few new things.
- Kernel 3.1x
- FirewallD
- SystemD
- mariaDB
- wait... someone saied in place upgrade.

There are dozens of other changes but this ones cought my attention.
I use Linux since years and every time I need to switch to a major version I know that I need to spend
a lot of time to reconfigure all the stuff and learn what's new to configure them on their best.

In place upgrade is a "miracle feature" that doesn't exist and that cannot work on a real life server.
I have spent 7 days to switch from CentOS 6.5 to CentOS 7, no automatic tool can do this automatically.

My server is quite simple, I have:
- Apache (PHP), some vhost, ipv6
- Tomcat (Java)
- MySQL
- iptables
- Various init.d scripts
- various cron
- fail2ban
- custom logrotate
- Postfix
- Sieve filtering
- Amavisd
- Dovecot
- Roundcubemail
- A multiplexer to "workaround" hotel/company firewalls while offering service from 443
- snmpd
- Cacti for statistics + mailgraph and pflogsumm
- SVN server
- OpenSSL for web servers and for emails

No tools can port this software on centos 7.

Kernel 3.1 is nothing more than the core that feeds all the new features, its head is Linus Torvalds, so no need to argue on it,
one of the best piece of software ever written.

Let's start analyzing the other new cool features.

SystemD is so fast and so asynchronous that it breaks more than it fix.
Changing scripts from initd to systemd is quite easy and it requires small effort.
The real problem is in the asynchronous start.
It took me a day to understand why postfix doesn't start with systemd while IPV6 networking is enabled.
In few words postfix started before the system got an IP and this breaked all.
Systemctl is no way good as ntsysv, ntsysv give you all the info you need with small effort,
it gives you what services are present, what are enabled, what are disabled and give you info on the service by pressing F1.
systemctl bla bla bla bla to do all this things

systemctl start postfix.service
service postfix start

I absolutely prefer the initd style.

FirewallD refresh the way you use iptables and it's good and bad at the same time.
It is useful if using IPV4+IPV6 since it can manage both IP version with the same configuration, no needs for two config files and two different rules.
The bad is that you need to study a new style to do always the same things.
FirewallD brings "the concept of zones", ok, I like it on my windows pc, I don't need it on my server.

EL7 repository lacks too much packages, EPEL is quite empty and this is not good.

PHP is compiled with a more restrictive policy, it blocks the use of TLS socket while using self signed certificate.

The switch from MySQL to MariaDB is completely painless.

Never spent so much time to switch from a major version to another, in place upgrade? ahahah

screwballl
Posts: 8
Joined: 2014/07/18 18:45:53

Re: My two cents on "the new Linux" -> RedHat EL7 / CentOS 7

Post by screwballl » 2014/07/18 19:37:25

I installed it on a test/development server and running into all the same problems as you are. We plan to stick with Scientific Linux 6.5 which should have long term support for at least another 5-7 years.

jmrcpn
Posts: 5
Joined: 2014/07/18 19:06:00

Re: My two cents on "the new Linux" -> RedHat EL7 / CentOS 7

Post by jmrcpn » 2014/07/18 20:00:00

You comment about systemd is not a surpise to me, I was expecting the very kind of trouble
you are reporting. Now, wait for a very real init problem, that to be resolved under
time presure in a production system....a real mess

Systemd very existence is why I won't put centos-7 in production.

By design, Systemd is trying to do everything to get "total control", this is a good strategy with a closed system (a black box).
AFIK, the unix/linux success came from the fact it is (was?) an open system.

What surprised me the most about exchange on the subject are comment like this.
Remark: "Systemd is very wrong by design"
Answer: "Yes there is few bugs, but we are fixing them".
Hope you see the irony.

Systemd is to initd , what Concorde is to B-747, or Multics to Unix (I am not fair with Multics as it came before Unix).
Computer guy are not the only one reaching the "complexity wall". Commercial flight pilotes reach it too, seems there is more
and more concern about pilots not anymore able to cope with the unexpected.
Flight AF447 was terminated in 2 minutes chrono because missing velocity data, while 20 or 30 years ago, in the same
situation, pilot would overcome, it would have been only a tense situation.

"unexpected" is not well managed by fully integrated system as Systemd, the Systemd challenge is to get rid of
the unexpected, grabbing more and more responsabilities overs applications, becoming, de facto, a kernel layer over the linux kernel.

sblantipodi
Posts: 252
Joined: 2009/07/10 09:43:13
Contact:

Re: My two cents on "the new Linux" -> RedHat EL7 / CentOS 7

Post by sblantipodi » 2014/07/20 00:32:46

It seems that I'm not the only one who don't like this upgrade too much.
I think that I need time but I will start liking it

_ck_
Posts: 89
Joined: 2012/08/10 23:00:35

Re: My two cents on "the new Linux" -> RedHat EL7 / CentOS 7

Post by _ck_ » 2014/07/20 23:19:23

Parts of centos7 can be disabled and added to maintain 6.5 compatibility, net-tools, rc.local, disable firewalld, eth0 naming etc.

But things like systemd and grub2, wow what a mess.

Systemd is the worst thing, hands down, to happen to linux, it is not even about getting used to change, it is just horrible.

Everything is 100% proprietary about it and for many things you have to remember to run commands after you make changes.

What it claims for simplicity is actually a huge mess of little files all over the system, ie. creating tmpfs directories for services, whole other area than the .service file and completely proprietary format you will never remember.

Grub2 is not any better, same thing with running commands after changes - just try to confirm after a kernel update what the boot sequence is going to be on next boot.

The problem is 6.x won't get anything but security updates after 2017 which is only 2.5 years away, which goes too quickly, then it is all over in 2020

User avatar
vonskippy
Posts: 839
Joined: 2006/12/30 03:00:04
Location: Western Slope Colorado

Re: My two cents on "the new Linux" -> RedHat EL7 / CentOS 7

Post by vonskippy » 2014/07/21 01:34:41

Systemd is to linux as the Windows 8 UI is to Windows.

In other words, a unusable cluster f**k.

How it was possible for a couple of Fedora flunkies to drive such an important system wide decision I'll never understand.

Huge mistake RedHat - Don't do a "Microsoft" and ignore the feedback from your users.
For the 2.5^15th time :: Better Details = Better Answers

sblantipodi
Posts: 252
Joined: 2009/07/10 09:43:13
Contact:

Re: My two cents on "the new Linux" -> RedHat EL7 / CentOS 7

Post by sblantipodi » 2014/07/21 18:06:50

vonskippy wrote:Systemd is to linux as the Windows 8 UI is to Windows.

In other words, a unusable cluster f**k.

How it was possible for a couple of Fedora flunkies to drive such an important system wide decision I'll never understand.

Huge mistake RedHat - Don't do a "Microsoft" and ignore the feedback from your users.
every time I read a post like this I know that is written by vonskippy.
as usual, your comments don't bring anything to the discussions.

dwaltz
Posts: 8
Joined: 2014/01/27 09:04:19

Re: My two cents on "the new Linux" -> RedHat EL7 / CentOS 7

Post by dwaltz » 2014/07/22 21:39:49

sblantipodi wrote: I absolutely prefer the initd style.
We all face the need to learn to sort out the dependencies of the modules to start up, but I don't think a procedural script is superior to declarative configuration.
jmrcpn wrote: Systemd is to initd , what Concorde is to B-747
I would say Systemd is to initd , what functional programming is to precedural, what sql is to manging your database by hand.
Sometimes we do have to worry about execution plans, but most of the time we don't know how a query works behinds the curtain: we just get the data by stating what we want not how we want it to come out of the datafiles.

balky
Posts: 5
Joined: 2011/10/06 16:33:38

Re: My two cents on "the new Linux" -> RedHat EL7 / CentOS 7

Post by balky » 2014/07/23 13:13:34

From a desktop user perspective, gnome3 is big no-go for the moment...
I am not ever going to recommend or have this set up for any of my users...
It looks like windows 8.1 to start with...
All the systemd cr@P et al...
It is so dis-heartening to see CentOS hopping on the windows wannabe bandwagon...

We're already starting to test / try Arch Linux...

gerald_clark
Posts: 10642
Joined: 2005/08/05 15:19:54
Location: Northern Illinois, USA

Re: My two cents on "the new Linux" -> RedHat EL7 / CentOS 7

Post by gerald_clark » 2014/07/23 13:23:43

CentOS didn't hop an anything other than RHEL.
Red Hat made these decisions and there is nothing CentOS can do about it.
If any change is to be made it will have to be at the request of paying Red Hat subscribers.

Post Reply