Replace the basic distro

Will it actually be possible to replace the basic distro in future?
Scenario:

  1. Installation and operation on Rocky-Linux
  2. after some time, a backup of this is restored to a new Debian instance and put into operation
  3. the old installation is decommissioned
    Sincerely, Marko
1 Like

Yes backup/restore is possible, but you can also join another node with a different distro and move modules to it.

Bear in mind that mixing different distros in the same cluster is a bad idea, because it becomes difficult to handle OS updates consistently.

However it is still acceptable as a temporary bridge just to change the distro.

3 Likes

I can understand the underlying base strategy. However… if your project is meant to be available for several distros… will be a requirement for the cluster? Uniform distro across the nodes?

Anything else would make no sense. In my eyes…

Regards

Uwe

1 Like

In principle I completely agree that it is good to have a uniformed distro for all of the nodes.

However there is bound to be a certain situation where there is a requirement to run a different OS for some oddball edge-case technical reason.

I do agree to this too, but I really wish that people would learn their lesson, and the “chosen” distro wasn’t a RHEL clone!

Red Hat is not to be trusted, how many more issues do they have to throw at the people for the people to realize this?
Rocky and also Alma really do not relly have the resources to have their own EL development, it’s still downstream somehow from RH.

Oracle does have enough resources for this, but I would NOT use Oracle underneath, no matter what!
Oracle has blown away too much trust in the OpenSource community!
OpenOffice / MySQL / now JAVA…

At least Debian is supported for the basis, so I’ll just use Debian for my and my clients nodes…
(and innerly laugh at gullible people…) :slight_smile:

My 2 cents
Andy

1 Like

…me too.

It was more of a basic question out of interest.

2 Likes

Cheat me once, blame on you.
Cheat me twice, bame on me!

→ I’m too stupid to learn !!!

But RH / Big Blue has been doing it a little too often!

My 2 cents
Andy

3 Likes

And for production environments should be a golden rule…
Neverthless, a lot of software/packages come for several distros, NS8 decided to be distro-agnostic (with some limits due to availability of packages on every flavour).

With a working cluster (made from more than 3 servers) with the same underlying OS, for instance I’m willing to try a different distro (different support for storage or filesystem, another branch of kernel development, point release instead of rolling); NS8 should be there to allevate me the container hassle.
Install OS and core, adopt node, backup of a container from one node, restore to another. It’s still podman or docker, the same orchestrator behind… should work without quirks or adaptation.
Otherwise, distro-agnostic start to become quite pointless and only a way to broaden the installation base.

The question is for future answer. But the answer should be put into system requirements, top of documentation for careful deploy the cluster environment.

Yes it is: you can do it, if you want. Don’t loose control of your experiment or it can bite you on the next major distro upgrade.

Exactly. You are an excellent writer. Would you like writing a pull request for the docs?

I think it should be a good practice. As sysadmin, I don’t particularly appreciate learning and managing too many different linux distributions. In my experience, sysadmins prefer to standardise the basic distro in order to manage more effectively updates/customization and don’t get mad

Exactly, as simple as possible and as cumbersome as minimally necessary.

Regards…

Uwe

As stated before…

so

As a not-green-anymore person, willing to reduce effort and (supposed) time wasting knowing newer things? I tend to agree, for several reasons. However, it’s not the point.

If uniform underlying OS will be mandatory for clusters, IMVHO will complicate a lot the growth of installation base.
I’m expecting a migration path (dream zone… live migration? :smiley: ) from one node with one distro to another node with another distro (and @davidep stated that should be possible) can relief a lot of hassle for sysadmins migrating from community supported distro to a service level supported distro.

Because… NS7 mimic an OS. NS8 don’t.

I tend to agree. So basically the question(or wish) would be, how OS/vendor agnostic will NS8 be?