Are Linux Distros Redundant?

Home » CentOS » Are Linux Distros Redundant?
CentOS 25 Comments

I just realised that I haven’t touched a CentOS/redhat machine in more than a couple of years. Everything I do now is Kubernetes based or using cloud services (or k8s cloud services).

What about it listeroons? Is your fleet of CentOS boxes ever expanding or are you just taking care of a single java 6 jboss application that takes care of the companies widget stocks?

How are your jobs changing?

Cheers,

Andrew

25 thoughts on - Are Linux Distros Redundant?

  • What OS are your k8s clusters running on? How about your cloud providers? Mine are on RHEL and CentOS.

  • On top of that, I often use CentOS as my base image for Dockerfiles, using the CentOS docker repository. While I also use alpine sometimes if I want it really tiny, it is really nice to be able to just rebuild the image against updates and not have to worry as much about underlying API changes breaking my app.

  • If its old coreOS then it is a rebuild of Gentoo. If it is newer coreOS it is a repackaged Fedora/RHEL. At some point it may be a repackaged CentOS.. but it isn’t at a level you deal with. You have moved up the stack.. other people are dealing with the lower levels of plumbing and you are able to deal with higher level operations.

  • Some of us build the infrastructure others use, which is what Google does. That includes having baremetal servers people can access
    (reserve, tell it which OS to install) to develop code for PCI devices which will later be used by the Google and Amazons which will then abstract them to their users.

    If you are working at the kubernetes level, you could not care less about whether a given NIC works or is giving the maximum performance;
    you do not even care about which NIC is being used.

    It is all about what you do.

  • Andrew Holway wrote:

    Nope. Well… actually, my manager’s talking about Ubuntu or maybe even FreeBSD. He’s *extremely* upset with RH being so slow – 8 should have been out for some time, for one, and a lot of 7, even with SCL, is far behind, and our researchers want newer software.

    But most of our Office’s work is done in-house. Lots of CentOS and RH, lesser amounts, AFAIK, of ubuntu.

    Cloud? Why would I want to go back to time-sharing on a mainframe?*

    mark

    * Go ahead, explain the difference to me, and if you start to write “but it’s many servers”, then you don’t understand timesharing on a mainframe.

  • I’m wondering what desktops you run then, are they also running on Kubernetes? I know some prefer Windows or Mac OS, but others really like Linux to work with. How would that work if no Linux distributions exist anymore?

    Apart from that, there are people in this world who like to stay as far away from G**gle as possible. And there are some who do it with good reason and the same applies to A**zon, A**le, M$$rosoft you name them. They will never ask if distributions became redundant.

    I’m afraid too many clouds make the wider horizon invisible :-)

    Regards, Simon

  • Maybe you should try to explain to your manager why RHEL/CentOS exist and why it’s widely used in the corporate world. If he talks about Ubuntu then you could explain to him what Fedora is any why and how it differs from RHEL/CentOS.

    Of course, managers do not always listen to those who do the real work.

    Regards, Simon

  • I’m not really sure that the reasons for Rhel really exist anymore. The oft quoted Library stabilty is more of a hindrance than a help in modern development environments with well operating CI.

    When the dinosaur IBM bought RH it was clear that it had become a fossil.

    Of course there is still legacy applications that need that but I see a definite shift away from OS dependant monoliths even in the more traditional enterprises

  • Another point is that Ubuntu is not just a Fedora alternative, they have a long-term support option known as LTS – all the even numbered releases: 14.04 (at EOL), 16.04, 18.04 (latest). I have heard that for 18.04 forward, they are going to a 10-year support model. For a Fedora alternative the odd-numbered releases should be used.

  • Simon Matter via CentOS wrote:

    Bad assumption, in this case. My manager is also a sr. sysadmin. He pushed CentOS a long time ago – we were running 5 when I got here, almost 10
    years ago. But the folks we support keep wanting to run software that uses much newer PHP, and Python, and stuff from newer kernels.

    For example, allegedly (don’t know for sure), some version of Ubuntu supports CUDA out of the box, as opposed to the mess I have to go through getting it and updating it from NVidia.

    mark

  • Simon Matter via CentOS wrote:

    Btw, right now, we’ve just built a new server as Ubuntu, because my manager wants to use it to test zfs, including its ability to a) act as a RAID, directly, without an underlying RAID, and b) encrypt the whole thing natively.

    mark

  • I think that’s a very narrow view of what Red Hat does.  They’re not just writing rpm spec files and building somone else’s code.

    Red Hat is the largest contributor to most of the core GNU/Linux software stack.  If you use GNU/Linux, then you’re using Red Hat’s work.

  • Kubernetes isn’t really a general-purpose UNIX operating system and so the question seems like it’s comparing apples with oranges, at least, unless you’re doing a very narrow and specific thing with certain automation and scalability requirements.

    I don’t think the fundamental raison d’être for a UNIX workstation or server operating environment has changed in four decades, even if there are also specialized grids for scalable application or HPC operations.

    Cheers, Ben

    —

    Benjamin Hauger SysAdmin/CSDC-DMO
    Rm. 94
    x8371

  • I’ve been running ZoL on CentOS for years. Wonderful stuff. SysAdmin’s dream, although we keep all ZoL boxes off any public access and update on a carefully tested schedule to ensure that no RPM version weirdness happens.

  • I may only be guessing, but maybe he was referring to the age of C7. There is no doubt that C7 is now VERY old software wise. I’ve just gone through building a new web server, and have had to use a lot of external repositories in order to pull in even reasonably new PostgreSQL / Apache / PHP etc.

  • Perhaps you’ve never seen the Software Collections repositories?
    https://wiki.CentOS.org/AdditionalResources/Repositories/SCL

    For what its worth, RHEL8 (and subsequently, CentOS8) will have Application Streams, which will be somewhat like SCLs, except they’ll be more core to the OS. That’ll let you update software like python, perl, apache httpd, etc. without interferring with the OS.

    https://developers.redhat.com/blog/2018/11/15/rhel8-introducing-appstreams/

  • James B. Byrne via CentOS wrote:

    But, don’tcha know, the only way to clear the fog is to send lots of money to them….

    mark

  • Well, of course, because they still run it on hardware servers with CPUs, storage devices, networking, cooling, whatever. That’s so yesterday, the future is Serverless Computing… a bit like brainless thinking :-)

  • I know this much, Scientific Linux shutting down?….is a definite sign!
    (And I’ve loved SL from my first foray into Linix….in 2002!) sad to see my “personal favorite” as a server OS going away

  • technicaly SL is shutting down but in reality they are just merging with CentOS effort wise. this is not a bad thing as there was duplication of effort being that the outcome was so similar. so in effect the rhel based platform just got more focused. regards peter

    Sent with AquaMail for Android https://www.mobisystems.com/aqua-mail