Ubuntu as a supported distribution?

[thesilentman]

Hi John (and community), just one more thing

I am open to any opinions on this. Just proposing...

Have you considered offering DA support for Ubuntu as a standalone OS and not through being a derivative of Debian? I raised the point a few posts earlier, but it seems it fell of the radar.

My points are:

- Ubuntu is not "just" a derivative of Debian, but it's a major player in Openstack based clouds, pretty stable and always up-to-date. I would love to be able to run Directadmin in such a cloud

- Ubuntu lets you upgrade from one LTS version to another or even to intermediate versions. I am right now on Centos 5.9 and I've had it with only old releases of packages that are supported out of the box. Don't get me wrong. Centos is stable as a rock, but that alone doesn't really cut it anymore. Upgradeablility and flexibility are becoming more and more important, in addition to stability. And Ubuntu shines in that department. (Not wanting to start a flame war here guys )

- A major point in supporting Ubuntu itself and not as a derivative, is that Ubuntu has 2 releases a year and where the latest ones may already include stuff from a not-yet-released Debian version. That would make an upgrade, to such a version, not possible without risk, until the associated Debian version is out and DA supports it.

Cheers,
Frank

 

[nobaloney]

I've ,moved your post; it's now a new Feature Request.

Hopefully it will attract some discussion and DirectAdmin staff will be able to gauge interest.

Jeff

 

[thesilentman]

Great!
Thanks Jeff, lets see what happens

Cheers,
Frank

 

[Richard G]

I don't want to start a flame war either, just giving some nuance to a point or 2.

Centos is stable as a rock, but that alone doesn't really cut it anymore. Upgradeablility and flexibility are becoming more and more important, in addition to stability.

If you start with an end version (5.9) you can't expect updates to intermadiate versions anymore. If you started with 5.1 or so, you would have been able to upgrade all versions until 5.9 without problems.
However in the hostingworld upgradeability is not more (but indeed almost equal) important then stability (which the customer want's) and so that alone still does cut it, as is proven by the majority of the hosting world running Centos.
And it's not stability alone, also the yum package manager with it's exclusion functions which apt-get does not have, makes it easyer to use.
On the other hand sometimes it takes a bit too long for Centos to catch up in some cases, with upgradebility.

It's however very nice if Ubuntu is able to upgrade from let's say version 12.x to 13 without problems. That would really speak as a positive point for Ubuntu.

is that Ubuntu has 2 releases a year and where the latest ones may already include stuff from a not-yet-released Debian version.

Just as you could see Fedora as a kind of pre-release version of RH Enterprise and Centos editions. But Fedora is supported.
So seen in that context... why not Ubuntu indeed?

I won't use it, but I let's see, there might be a lot of interest out there for it. You have some good arguments!
Good luck.

 

[thesilentman]

Hi Richard, thanks for your input.. just a few notes from me

I don't want to start a flame war either, just giving some nuance to a point or 2.

If you start with an end version (5.9) you can't expect updates to intermadiate versions anymore. If you started with 5.1 or so, you would have been able to upgrade all versions until 5.9 without problems.

Of course, you're right with your point, Richard. But I started way back with 5.3 and I have updated all the way up to 5.9.
But some parts of the provided packages (even with EPEL or other repos) go only up to a certain point. For example db4 is stuck on 4.3 (db6 is current now). Perl is stuck on 5.8.8!!! (We re at 5.18 now, with lots of improvements) If for any reason you want something newer, you got to start doing manual work. I have no problem doing manual work if it's only for one thing, but if it's Perl, then db4 then something else, I find that annoying. It shouldn't be so hard to do in 2013!

However in the hostingworld upgradeability is not more (but indeed almost equal) important then stability (which the customer want's) and so that alone still does cut it, as is proven by the majority of the hosting world running Centos.

Agreed, that why I wrote "in addition to stability".
Concerning the "majority of the hosting world running Centos". Clearly that might be the case at this moment in time, but things are not standing still. As better technologies evolve people tend to chose what's best for them at the given time.

And it's not stability alone, also the yum package manager with it's exclusion functions which apt-get does not have, makes it easyer to use.

Actually, it has such functionality and even more... its called holding packages --> https://help.ubuntu.com/community/PinningHowto (You should also read the pinning howto. Very interesting )

On the other hand sometimes it takes a bit too long for Centos to catch up in some cases, with upgradebility.

The problem is that with Centos there is no upgradeability (see db4 and perl examples above).
In the Ubuntu world, you can for example chose to remain on a stable OS version while getting packages from newer versions. That is a huge flexibility factor.

It's however very nice if Ubuntu is able to upgrade from let's say version 12.x to 13 without problems. That would really speak as a positive point for Ubuntu.

Just as you could see Fedora as a kind of pre-release version of RH Enterprise and Centos editions. But Fedora is supported.
So seen in that context... why not Ubuntu indeed?

I won't use it, but I let's see, there might be a lot of interest out there for it. You have some good arguments!
Good luck.

Thank you very much Richard, let's see

 

[Richard G]

I see in your new reply, you explained a lot more and we agree a lot there. Especially these kind of things:

The problem is that with Centos there is no upgradeability (see db4 and perl examples above).

Centos has some upgradeability, but as I also stated, it takes a bit too long for Centos to catch up. I've seen newer things in Fedora (like Glibc versions) which did not even got to the same version in Centos almost a year later. That is indeed way too long. And I fully agree this should not be needed anymore in 2013.

Thank you for the link about the pinning article. I'm curious if that's also in Debian, because I ask a couple of years ago, and nobody give me a clue on howto exclude for example mysql from updating when updating OS packages. So it's very good that Ubuntu has such thing, and I'm interested so I'm going to read it, thank you.
I might even going to test Ubuntu sometime on a server at home to see which progress has been made.

Thank you very much Richard, let's see

Thank you to for this nice clarification! Hope you will get some votes!

 

[nobaloney]

I hope, Richard, that after you learn about pinning, you'll let us know if it's as easy as the concept of the yum exclude list, which to me makes yum a no-brainer for a system such as DirectAdmin, which updates so much of the hosting stack independently.

CentOS of course is a free rebuild of the Red Hat upstream, which is supported by Enterprise buyers who know that the stability they want is worth the price.

I and a few of my friends had actually started to develop our own Linux Distribution aimed at the webhosting industry, but we gave up when others started rebuilding RHEL; after all our business is hosting, not building OS distributions.

Jeff

 

[thesilentman]

Hi Jeff,

When you mentioned "pinning" in you post I think you actually meant "holding", because "pinning" is a mechanism that RHEL or its forks unfortunately don't have.

"Pinning" gives you the choice to install packages from a newer distribution, if you have need for that.

Besides Pinning, one of the other great advantages of Ubuntu, for me, is the upgradeability.
With RHEL/Centos you can't (without risk and hassle) upgrade from any 5.x to 6.x. Even Redhat says, you shouldn't and you do it at your own risk. With Ubuntu, it's possible.


BTW, the equivalent of yum's exclude commands/list is:

apt-mark hold packagename1 packagename2 etc..

Easy

see: http://manpages.ubuntu.com/manpages/precise/man8/apt-mark.8.html


which wraps around dpkg --set-selections

see: http://manpages.ubuntu.com/manpages/precise/man1/dpkg.1.html



Just a small note on the side. I have worked with Centos for a few years and I found it most stable, but I have raised my reasons for liking Ubuntu more during this thread. And Ubuntu LTS is targeted at the Enterprise. In the cloudspace many large Enterprise use and prefer Ubuntu over RHEL/Centos.

Greetings,
Frank

 

[Richard G]

@Jeff: I had a short look but I still find the yum exclude list a little nicer to work with.

apt-mark hold packagename1 packagename2 etc..

Yes and you can show what's hold by using apt-mark showhold.

I just wonder if it has also the ability to look for x64 if present and if not use i386 instead.
Because in Centos you can exclude i386 packages, but you can also use the line "multilib_policy=best".

Personally I would like it better to work with configfiles like yum.conf, because you can see what's excluded and directly see if you made a typo and easily change that.
If you want to do that with pinning, forgive me if I'm wrong, but then you first have to unhold the package you made the typo in and then hold it again, taking care you don't make any typo's again.
Unless this can be fixed with the -f=filename option.

The pinning method might have some benefits, but it's too much commandline imho, yum is really a no-brainer and a little easy configuration file so I still prefer yum from what I've read.

 

[nobaloney]

Thanks, Richard. At this point I still side with you and the long history of Red Hat (and therefore also CentOS, built; from the same source code). for others, well your mileage may certainly vary, and I'm looking into offering various OS versions in our forthcoming VP servers.

Jeff