Future Proof of your hosing business by Choosing Debian or Ubuntu

DanielP

Verified User
Joined
Jun 28, 2019
Messages
222
Red Hat Enterprise Linux (RHEL), offers several advantages that make it a popular choice for businesses and organizations. RHEL is known for its stability. Red Hat provides long-term support for its releases, which includes security updates and bug fixes without major changes to the system, making it ideal for environments where uptime and reliability are critical. Long-term support, and enterprise-grade features of Red Hat Enterprise Linux (RHEL) are indeed significant reasons behind the creation of RHEL clones and why it is chosen 00 to be the main distros of the Hosting Indursty...

How ever main clients of the RedHat are Large Enterprises and Government bodies they usually buy certified servers and run on them for the entire lifetime (do amortizations) then throw them away and by the new one repeat the process. they are ok the performance remain the same trough the entire lifecycle of OS+Server

clients.pnghttps://www.redhat.com/en/success-stories


We are different I recently did yabs benchmark on an old 2670 v2 to evaluate connectivity for location been used as backups it also runs geek bench6... 10 years ago such server will be considered pretty decent performance
But if we ran it like a RedHad customer with current day performance on geekbench6 single tread of 650 which is important for PHP last couple of years that CPU will be classified as unusable as 5 years old 3900x 3800 Ryzens has 1600 59XX has 2100 and new ones 7950 and EPYC 4004 are in 2800-3000 zone ...

So for a hosting provider to stay competitive we got to change hardware more frequently ... During the 00 periods and the 10 when Intel dominates the CPU market that is not that much of an issue as performance do not changed that much during several generations of CPUs especially for E3 family where 1245v2 performs not so differently from 1230 v6 which on the single thread benchmark (with in the 10%)



but now days jumps are bigger to workaround that I decide we will run on virtualized hardware we own hardware in one of our locations but we rent in others and that is even better as we can easily migrate to new one on even shorter cycle


and here starts the issues

Red Hat designed their linux for their paying clients needs the product lifecycle is that

life.png






chart-life.png

Native hardware enablement is provided by backporting hardware drivers, etc., to the relevant version of Red Hat Enterprise Linux. Hardware enablement using virtualization is achieved by running an earlier version of Red Hat Enterprise Linux as a virtual guest on a newer version of Red Hat Enterprise Linux.

Native support is for the first 5 years of full support

I first notice the issue when in early 2020 Ryzen 3900 become modern with some hosting companies and we also got some I migrated some CL7 VMs to such hypervisor nodes and it booted with warning for unrecognized or unsupported cpu.... Well x86 has great backwards compatibility and it worked BUT I deviated from RedHat / RH Clones path by providing such new CPU to old version and in that moment any strange issue that happened on the server - can be blamed of running on such unsupported cpu ...

Later Cloud Linux as a great company provided hybrid kernel 4.18 for CL7 with which CPU become recognized and it was stable with it too... but that is again another deviation it is not the RH Clone with guaranteed stability ... with same success CL can provide Ubuntu kernel for CL7 - it is something new with its own stability... not 1:1 RH Clone

now it is 2024 Native hardware support for new hardware just ended couple of months ago for RH8 and it's clones that are 1:1 binary compatible - for any CPU released after that date will gave (We expect 9XXX series Ryezen soon)

That mean for if I want to keep it on RH8 Clone path for the same guaranteed stability I cannot use anything than hardware released before March 2024 - or went next hybrid kernel - path deviating from 1:1 RH Clone path...

and in 3 years native hardware support for new hardware ending for 9

============================================================================

When RH and its has been chosen for it's stability and LTS only RH offered such LTS... Ubuntu was 1 year .10 versions was like a dev one - you can skip upgrade for it .04 has support until next .04 (I started Ubuntu since version 8 as desktop) that is how I remember it and do not used Debian initially until couple of years later BUT now both Debian and Ubuntu have lts and it is battle proven fist by us as we run Proxmox which is Debian with Ubuntu kernel and second by the industry as majority of open stack clouds run on ubuntu so if you run virtual RH Clone for it's stability in a lot of cases you relay on Ubuntu stability

============================================================================

When CL Released version 8 ready for production in 2020 I adopted it and did some test there was 10-15% performance difference 7 vs 8 where 8 is faster... BUT adding we run on vitualized have some additional overhead ... for my user case is beneficial to have easy migration between version of OS which runs newer components

RH Clients do not need that so 1:1 clones do not have that capability ...

Also that Brings question for OS version EOL...

Sure Alma and Cloud Linux released Elevate but does someone notice the docs CL one says

I have a CloudLinux 7 system with DirectAdmin/Plesk/another panel installed, how do I upgrade to CloudLinux 8?

Unfortunately, CloudLinux ELevate doesn’t support these system configurations yet.
Instead, you can create a new machine with CloudLinux 8 and migrate your system’s license and configuration to it.
So a Conclusion of this point here instead of deviatiing - running on unsupported hardware

I did a lot of thinking did I try to run totally unsupported no panel elevate and see what will happened do i bug fix it but decide against that and did migration of the last of mine CL servers to 9

SO we (all of us here) were skipped ! on that production cycle and elevate way from perfect I ran in on my last cpanel server it took 3 hours -of which 2 + hours downtime and has couple of scary moments when it create in memory boot os with packets downloaded and start to uinstall rpms from disk before replacing them with new one... totally inefficient ...

Debian and Ubuntu Solve these replacing the name in deb repo for Debian and running upgrade and dist upgrade took couple of minutes for 10 to 11 and for 11 to 12 and I then did build all and do-release-upgrade 20.04 to 22.04 took 15

then it is simple da build all - for just in case you are good to go ...


Last But not Least RH Dead Cycle:

that is something i first notice that in 2017 when around RH 5 EOL where 2 companies in a distance of couple of months sold themselves one was the oldest on that local market (Market Player 3) by size and the other is very reputable one (Market Player 5) both was bough by Market Player 1 ... for which ex owner I think of as cheapskate .... any way in 2021 early just after centos 6 eol Market Player 1 announced they bough market player 2 (market player 2 was not using cloud Linux but some old tools from h1 in the past but they never migrate to CL so it was EOL for them and big migration ahead

this is not some conspiracy theory because the ex market player 4 which is now number 2 on that market - was my friend from before our web group create a hosting business... he approach me couple ocassions and try to sales peach me how hard it is tor run a biz ... what heavy migrations are coming and was wanting to buy me .... not knowing that I prepare for the dead cycle and got only 4 cpanel servers to migrate and 2 DA ones from 7 to 8 ... for which attempt because was too pushy I now classify him as a frenemy ...

So no ahead is preparation for 2029 - even I like CL I'm currently 40% Debian and Ubuntu and will push that percent forward because all issues I mentioned are solved easily if you run your business on OS that can easily upgrade to next version.... cutomers on a server can be without migration literally for ever... , just jumping hardware and os version ... and when current size storage become obsolete I will just add new virtual hdd and move home and mysql to it keeping / thin provisioned for the os only...



That is for today :) I hope you enjoy it
 
Last edited:
It is a tedious and complex task for us (and DA probably) to follow those software , hardware patch/update/upgrade.

---

We developed our own scripts (using DA admin backup/transfer) with rsync(s) to migrate users from one server to another. (mostly within a hour for xxx GB data).

---

However, to keep the business continue, we need to adopt those continuous changes.
(e.g. OS update/upgrade, DA update, Apache/Exim/Dovecot/database update/upgrade, new PHP released, some old stuffs obsoleted, etc etc, test it before live).

Nevertheless, IMHO, the hosting marketing seems shrinking.
 
@ccto I spend yesterday reading trough credit ratings of webpros which make cPanel and Plesk ... they have such because they have a large debt in those credit ratings both Fitch Ratings and S&P Global agree on that hosing market is growing mid to high single digit per year (5-7%) and is expected to do so next couple of years...

@Richard G @pat0 @gate2vn - here my thread - I finished it yesterday and just post it.... today I fixed some typos and I'm tagging you as I promised ... I would like to hear your opinion about what I wrote if you have such ...
 
Last edited:
I think @Active8 and some others did some elevating from Alma 7 to 8 and from 8 to 9 without issues and without that much downtime.
And also using ofcourse the build all command afterwards because that is required for things to run smoothly again.
At least it's better than which we had before, because there was never any elevate stuff before and if you needed to upgrade from Centos 5 to 6 for example, you either had to get a new server and move everything, or make a backup, empty the server and install it with new OS.

But yes it's worrying, but to me not a lot of worries, since we don't use any virtualizing anyway and switch to other servers all 5/6 years.
So if things are tending to go sideways, we can always switch to Debian or Ubuntu.

Question is if we will be doing that due to what is happening to legacy licensing. Since we're at a nice age anyway, it might even be that we are going to stop our business in like 2026 or 2027.
 
All our servers are migrated from Centos 7 and/or AL 8 to AL9 by using Elevate script.
We didn't encounter any problems maybe because we don't have fancy server setup, just plain setup.

Total time was max 1 hour incl. system updates (AMD Epyc CPU) .
For now we are not considering Debian flavor , some servers have Ubuntu distros (test machines) and works well but nothing in production
 
Thanks for sharing your thoughts @DanielP :) You made me feel old. Our first servers were running Pentium 4 :D Then upgrading to Opteron dual cores... ways too long back :D

For the success stories, well... I believe that any brand would like to list "big names" as their customers, to make them look good. So, while RH has big companies as their names, definitely, there are other smaller customers too.

For the ELevate script, the problem is not the OS itself, but the applications installed by control panels. If you run ELevate, it will give you a list of incompatible software, which you can fix before or after the switch. I did several times, from CentOS / CL 7 to Almalinux / CL 8, 9. There are downtime when rebooting, but not three hours in my cases. RH, Ubuntu or Debian, when it installs a new kernel, it always needs to compile, and there is always a risk at that step, isn't it?

About the CPU or new hardware, as a hosting provider focusing on stability, I am not ordering the newest hardware immediately when it's released. The new CPU seems to be attractive, but it takes time to have compatible software. For me, hardware technologies are always running ahead software. Many of us are virtualizing our hardware, then it even becomes more complicated. So for me, I am not worry about the hardware support, but more about the stability, the security of the software.
 
Here in screenshots how elevate went for me - CloudLinux 7 cPanel elevate
when it booted in ram os i watched it with vnc and screenshoted my laptop clock which is one hour ahead from server

(I do partial screenshots from the original screenshots for the post as I usually screenshot entire screen)

VM is 2 x 16 cores 2690v4 128GB RAM at that moment

start time is 0:39 server time say it 0:40

elevate01.png

Around 1:15 it stoped services - server is online but do not server client pages

elevate02.png

10 minutes later
elevate03.png

1:31 50 minutes into process I decided to open second session and checke where am I - look good 3 of 5
elevate04.png

1 hour 5 minute into process - 30 minutes downtime for the clients - looks very good it says rebootin into stage 4

elevate05.png

3 miutes later 1:49 server time, 2:49 laptop time 1 hour into process - sill look promising
elevate06.png

2:01 server time 3:01 laptop not a speedster but 1400 of 4600 yum updates I know half are cleanups

1 hour 21 minutes into process 44 minutes no service for the customers -

elevate07.png



17 minutes later 2400 of 4628 - here it scares me a bit - almost in the middle where I expected to start cleanup
I start displaying once a minute no dev log not moving process list

elevate08.png



6 minutes - later - finally it start to cleanup make me breath easily that process not crashed ...

elevate09.png


2:35 server time 3:35 on my laptop it finish clean up and do some movements
1 hour 55 minutes since process start up - 1hour 20 minutes since customers do not have service


elevate10.png



10 sceeenshots are allowed per post so I will continue in second post
 
5 minutes later it says report is generated

elevate11.png



and rebooted - Im out of danger part - VNC shows it will boot EL8 Kernel make me smile

2 hours into process - 1 hour 27 minutes no service for the clients

elevate12.png



Well it is too soon to be happy even it is booted into EL8 kernel - that's all scary ram os boot and rpm relacement was not stage 4 but just preparation for it or the first half

2:42 server time - Starting stage 4 - here I lost my smile

elevate13.png


20 minutes later 3:06 server time still doing some things 2 hours 26 minutes into the process customers do not have service for 1 hour 51 minutes

elevate14.png


3:26 Some minutes ago customers received service after about 2 hours downtime - mysql litespeed started - but cpanel is still downloading


elevate15.png


10 minutes later 3:35 server time 2 hours 55 into the Elevate process it says rebooting into stage5

elevate16.png


and this was the moment for the final joke of the Elevate process after the reboot there is no ping :)


Well I logged trough VNC is says stage 5 completed - when I did some check name of the network card is now different inside of the VM :)


So it was and easy fix but but took me couple of minutes

I will not repeat that 3 hours - I can move even 200 accounts faster ... with DA advanced move

I forgot to take last screenshot - but not care

So that is how elevate happens for me on May 26 2024
 
Last edited:
Long time indeed. But what is this, an external tool or something which is in the modern licences nowadays?

sorry misspelled that what I wanted to say is
C. (Advanced) Partial backups with home and mysql sync - from the migration manual

That is the best and the fastest for full server migration if you do 1:1 in my experience
 
I can see why you don't like it :) But again, it's not the OS itself. The process started with EA4 profile, and during the upgrade, many times I see ea-packages. That is cPanel related. Unfortunately, I don't have any cPanel server for playing with.
 
Choosing between Debian and Ubuntu depends on what you need to future-proof your hosting business. Both are solid choices, but they have their strengths. Debian is known for its stability and is perfect if you want something rock-solid that rarely changes.
 
Back
Top