Does openindiana, sco-unix, unixware has features freebsd does not have.

Some make money by selling "systemd-services". Because,well it's complicated.
I still complain about systemd at $JOB and have been told learn to like it.

The real problem with systemd is it's opaque. There are something like 40 or more daemons associated with it. Certainly something this complex is not only difficult for people to use but difficult for its developers to maintain. We all know the saying, complexity is the enemy of security.
 
The real problem with systemd is it's opaque.
I had an issue one time related to USB stick mounting and having an entry in fstab. System stopped booting, some stupid windows-like error code number on the console that had to go to another system and google grep "wtf does this mean" to find out "use this tool to look at binary logs and find out" "oh, usb, fstab entry"

Learn to like it? heck no. Learn to get by? Ok if you pay me enough
 
I never complained, neither did any other Solaris or any other UNIX admin complain.

But you wrote:

BTW, IMO the Solaris SMF was the correct way to solve the problem.

You are speaking about solving a Problem that is not a Problem. And why to form FreeBSD as the ones that
complain and should use Linux, because that is their reference point? Desktop freaks should use Linux,
Ubuntu, not FreeBSD.

I installed once Illumos, and got crazy with SMF. Also with the rc system of Linux. And also the one of FreeBSD
is too complicated for my taste. At best is the one of OpenBSD.

I think SMF is for people managing a lot of computers. For people with one or two servers, for people
using it as desktop, for embedded devices is not an advantage, it makes him life unnecessarily complicated.
 
Even though systemd is technically not the best solution, a lot of,many,linux distro's have systemd as default.
Some big players like redhat are pushing it
This is no wonder, because Redhat was Poettering's employer when he gifted cursed the Linux world with systemd.

I mean with commercial driven unices it's a common topic that all have replaced SYSV Init. Ubuntu had upstart, Apple has launchd, Solaris SMF. All are trying to solve more or less the same problem(s).
 
But you wrote:



You are speaking about solving a Problem that is not a Problem. And why to form FreeBSD as the ones that
complain and should use Linux, because that is their reference point? Desktop freaks should use Linux,
Ubuntu, not FreeBSD.

I installed once Illumos, and got crazy with SMF. Also with the rc system of Linux. And also the one of FreeBSD
is too complicated for my taste. At best is the one of OpenBSD.

I think SMF is for people managing a lot of computers. For people with one or two servers, for people
using it as desktop, for embedded devices is not an advantage, it makes him life unnecessarily complicated.
Don't twist my words.

1. "problem" as in the "problem" desktop types were talking about at the time. Again, I and my team never complained but there were a lot of amateurs on the various mailing lists playing with Linux who complained... And then the marketing types who defined "problem" got involved. I've been to a lot of meetings hosted by computer vendors discussing the "problem."

2. "Installed once Illumos." -- During my 45+ year career of which I've spent about 25 of that on Solaris and other UNIX variants such as Linux and FreeBSD, professionally. I've installed many operating systems from IBM mainframe operating systems to various UNIX variants. Not just once but hundreds of times.

3. "I think SMF is for people managing a lot of computers." -- I currently manage about 1000 servers in two separate datacenters in two cities. Does that qualify as managing a lot of computers?
 
This is no wonder, because Redhat was Poettering's employer when he gifted cursed the Linux world with systemd.

I mean with commercial driven unices it's a common topic that all have replaced SYSV Init. Ubuntu had upstart, Apple has launchd, Solaris SMF. All are trying to solve more or less the same problem(s).
Pottering now works for M$.

 
Imagine UCL formatted SMF manifests. I believe Zones uses SMF natively too. That would effectively kill the need for all this configuration management goo in the ports tree (yes, that includes you too, docker), and streamline service scripting in base.
 
Imagine UCL formatted SMF manifests. I believe Zones uses SMF natively too. That would effectively kill the need for all this configuration management goo in the ports tree (yes, that includes you too, docker), and streamline service scripting in base.
Like jails, Zones are a kernel isolation and partitioning technology. The term Solaris Containers are the userland definitions that structure zones in userland in order to present the Solaris "jail" (zone) to the end-user. Think of zones as a kernel thing and containers the environment.

FreeBSD VNET jails are the FreeBSD equivalent to Zones. While Solaris Zones are created and managed using Container definitions, FreeBSD jails are managed through the jail(2) system call. The jail(8) command uses the jail(2) system call to create, manage, and destroy jails just as Solaris Containers manage Zones. One could, for example write one's own replacement for jail(8) command which might for example using a GUI or ncurses interface, calling the jail(2) system call directly.

Solaris Containers and Zones have a similar relationship.

If one is keen to learn about Solaris Zones the book Solaris Internals by Ricard McDougall and Jim Mauro is a good read. (Which BTW you will learn that the SLAB allocator in Solaris predated the Linux innovation.) Of course even the Solaris SLAB allocator isn't that revolutionary because the IBM MVS (mainframe) subpools (memory management) is pretty much the same concept. Generally you want to keep similarly sized kernel memory allocations near each other to reduce memory fragmentation. But this is off topic now.
 
If I freewheel.
In FreeBSD you could put some configs out of the /etc/rc.d scripts and put it in a JSON/YAML file.
To simulate SMF ?
 
Doesn't /etc/rc.d/* does the same thing as SMF manifests ? Just with sh-scripting & without XML-formating.
It's more like SYSV rc on steroids. Rather than the order be determined by Snn* and Knn*, the order is determined by metadata in each script itself. It's sorta like both.

SMF and systemd support parallel start/stop of services. Our rcorder(1) also supports the -p flag but our /etc/rc does not support parallel start/stop, though I'm working on a patch to implement parallel start/stop of services. The patch hasn't been tested because I have a lot of other things on my plate that take precedence.
 
Pottering now works for M$.

I know, when this became known I wrote an own thread about it here. :D So yes, Poettering and Microsoft deserve each other, because Poettering has this precious Microsoft mind set since he started his programming career. And finally he can use it to improve WSL!
 
FreeBSD VNET jails are the FreeBSD equivalent to Zones. While Solaris Zones are created and managed using Container definitions, FreeBSD jails are managed through the jail(2) system call. The jail(8) command uses the jail(2) system call to create, manage, and destroy jails just as Solaris Containers manage Zones. One could, for example write one's own replacement for jail(8) command which might for example using a GUI or ncurses interface, calling the jail(2) system call directly.

Didn't SUN get rid of the whole Solaris Containers nomenclature when they finished Zones? I believe zoneadmd() is what provides the native management layer in conjunction with svc.startd(), and svc.configd() (and subsequently the configuration repository) for Zones (from SMF). Writing something like that for jails would be fun challenge, although it'd probably require deep kernel level changes. Not sure if vendors would adopt such a drastic (but good) change to base.
 
Beastie7,

No. Containers is the term Oracle still uses. Containers is a term, not a technology. Zones are the technology. Containers is the term used for the wrapping put around zones for users. (I work at a shop that manages 1000 Linux servers, 200 [was over 600] Solaris servers, < 100 AIX, over 8000 Windows servers, 6 VMware clusters, and thousands of cloud instances. I was Solaris Team Lead for 15 years. We still use the term containers.)

Also, what do you mean by "finished Zones?"

We already have something like that with jail.conf. Though, anything can issue the jail(2) system call. Writing a different interface would be technically easy. It's deciding what that command line interface curses interface would look like. Personally, I think this should be a Web interface. One could have a Vsphere-like interface, instead of managing VMs, managing jails. Or managing bhyve VMs and jails through a single interface. Certainly not part of FreeBSD but an add-on (port/pkg).
 
Also, what do you mean by "finished Zones?"

Everything you see in the lower parts of this diagram was developed in tandem with SMF, Dtrace, ZFS, etc; hence the "finished Zones" part. I was under the impression Solaris "Containers" was old nomenclature due to them (SUN) making the "Jails" idea (aka. "containers") more functional for the datacenter; thus Zones, or Jails++ if you will. ? The Zones text interface is so well laid out, I don't think a web GUI is needed. IMO
 

Attachments

  • serv-con-zones.gif
    serv-con-zones.gif
    75.8 KB · Views: 181
Back
Top