User Tools

Site Tools


moving_from_rhel_to_centos_or_oracle_linux

Disclaimer

This article should be considered an opinionated editorial rather than any type of unbiased review. As a potential Red Hat refugee, it'll give you a lot of useful info. However, It's not supposed to be an impartial product review of Red Hat Enterprise Linux. The author's opinions are merely his own views. There is little love lost for Red Hat here. Let me say that we love Linux but necessarily Red Hat due to bad experiences with their business practices, technical decisions, and product feedback cycle. We are not impartial.

Moving from Redhat Enterprise Linux (RHEL) to CentOS or Oracle Enterprise Linux

This page is dedicated to providing migration information for folks who have decided to move away from Red Hat and over to another Linux distribution. First of all, you might want to know that this is a common sentiment mainly due to choices made by Red Hat themselves.

I will use the terminology RHEL and OEL a lot. Those are short for Red Hat Enterprise Linux and Oracle Enterprise Linux.

Reasons to Migrate

These are the leading reasons why you might want to make a move away from Red Hat which have to do with Red Hat themselves.

  1. Red Hat requires every install have a subscription no matter what it's used for (dev, test, UAT, doesn't matter)
  2. Many believe Red Hat charges too much money for support. At this time, they will charge you $300 for a subscription to Red Hat and one can't even call their support at that level. You'll probably get “maybe we will respond if it's convenient” email support from their offshore teams. It's my opinion that level of support is not worth $30 much less $300 and customers feel squeezed.
  3. Many of the useful and interesting features such as clustering, glusterfs, and virtualization have been broken into separate and expensive layered products. Most sysadmins feel this defeats some of the reason to run Linux in the first place, turning it into an expensive walled garden.
  4. Red Hat's kernel patches require a reboot. Oracle Linux with ksplice doesn't have this issue (but it's a pay-only $$$ feature in OEL, just FYI).
  5. Red Hat began the move to Lennart Pottering's Systemd init replacement. For some, that move has been a painful disaster that has made administration more difficult, driven complexity up, and forced application management into problematic directions. For many, the move was an unacceptable move backwards.
  6. Some customers dislike Red Hat's update schedule, considering it to be overly frequent and aggressive.
  7. Patches and even trivial packages aren't available unless your subscription is up to date. If not, you have a partially broken system because you can't install even simple things like timezone updates or small unix CLI tools.
  8. Red Hat sales people have been known to act in ways customers can feel is too aggressive.
  9. Some consider Red Hat training practices to be overly monetized. For example, it appears very likely that they make up their RHCE and RHCSA tests with “gotcha” questions that you'll only learn if you take their $2500 - $3000 dollar classes, but those factoids are only to keep regular knowledgeable sysadmins (who learned from on-job-training) from passing their $250 test without paying the trollbridge fee for the classes.
  10. Red Hat has been bought by IBM who some suspect to have begun the process of strip mining Red Hat intellectual property and moving their staffing offshore. Many feel quality is dropping. For others, IBM's lack of profitability and recent 22-quarter losing streak also has shaken faith in the company in some opinions.
  11. Much criticism has been directed at Red Hat for making poor technical choices. Example: they eschewed XFS and badmouthed XFS for years refusing to support it, then started using it as the default in RHEL7 enigmatically. Now they applaud it.
  12. RHEL follows Fedora and thus any bad decisions in Fedora (and most sysadmins would never choose a desktop distro like Fedora that has a long history of controversial decisions) filter into RHEL eventually. Fedora are the folks who pushed first on “great” ideas like: systemd, dbus, Pulseaudio, XML config files, udev, python scripts instead of C-based tools, GNOME3, and other “replacements” which many system operators consider to be much worse than the thing they replaced.
  13. Red Hat isn't excited about BTRFS because the main developer works at Oracle. This could lead to more stagnation with Red Hat's already stagnant choices for filesystems. It's missing inline compression and deduplication without either BTRFS, ZFS-on-Linux, or ReiserFS v4.x, then it's really got no good option for getting any of those nice features.

These are the reasons folks usually migrate when they are facing internal struggles.

  1. It becomes too expensive to keep dev/test/UAT/training machines licensed with a subscription.
  2. The sysadmin gets tired of being cut off from OS packages and patches every time entitlements expire.
  3. Sysadmin frustration with the licensing portal or “Red Hat Satellite” package caches which are simply there for to prevent Red Hat being able to cut you off from your package repository any time they please. They still do not prevent Red Hat cutting you off from updates.
  4. The management gets tired of some of Red Hat's salespeople and their sales tactics.
  5. The customer needs clustering, ksplice, storage or OS virtualization, containerization or other features Red Hat can't provide, does poorly, or overcharges for.

Deciding on a Migration Target

The main way to decide on a migration target is to consider the reason why you want to ditch Red Hat. If you plan to keep Red Hat's tools and structure but you are just tired of paying so much, then CentOS and Oracle Enterprise Linux (OEL) would both solve your problems.

If you are completely tired of Red Hat (and as an RHCE I totally understand) and want to switch to another Linux distribution that offers your more for less and doesn't follow Fedora, then I'd suggest trying a Debian or Devuan based Linux distribution. These are more significantly different from Red Hat's ways of doing things. Folks just sick of Red Hat's methods and mentality can jump ship to any number of distros and be happy. If you want to abandon Linux completely, then perhaps consider one of the BSD distributions such as FreeBSD.

Check out the features below which are the ones our customers cite the most as being in play when considering a Red Hat migration.

Feature Red Hat Enterprise Linux CentOS Oracle Enterprise Linux Debian and Devuan FreeBSD
Package Access Only for subscribers Free Free Free Free
Patch Access Only for subscribers Free Free Free Free
Virtualization Basic support in RHEL, best features in RHEV which costs big $$$ Basic Features Basic Features Better Basic Features Basic Features
Reboot-Free Updates (ksplice) No No Yes for $$$ No No
Advanced Filesystems Unstable BTRFS preview in RHEL7. Nothing in RHEL8. Basically Zero. Same as RHEL Same as RHEL BTRFS and ZFS-on-Linux Full ZFS
Package Cache Hassle Yes, such as Red Hat Satellite or Spacewalk No No No No
Clustering Support Needs “High Availability” layered product for $$$ Basic Pushes you to Oracle RAC / Clusterware Basic Basic
Provides Upgrade Path Between Major Revisions Buggy and Problematic, but yes, barely. Same as RHEL Unreliable Oracle Yum Upgrade white-knuckled APT upgrade path Excellent “freebsd-update”

So, consider these scenarios:

  1. You are tired of every system needing a subscription even to have basic package functionality. Solution: Just update to any other Linux distro (or BSD) from Red Hat except for SuSE Enterprise (which has some of the same problems).
  2. You want a way to escape from Redhat without doing a huge dump-and-load migration on every box. Solution: Upgrade to Oracle Enterprise Linux. They have a tool that will convert Red Hat machines into OEL machines without having to completely wipe out the OS and start over. It's quite slick and it's existence infuriates Red Hat because they know it's a very easy way to abandon them.
  3. You want better choices for advanced filesystems. Solution: Move to Debian, Devuan, Ubuntu Server, or FreeBSD and use either ZFS-on-Linux (or native ZFS in FreeBSD) or migrate to BTRFS when it stabilizes. Also, see if you might be able to get a specific feature by combining LVM2 features with other filesystems. For example, Reiser4 has compression support and LVM2 has thin provisioning and advanced caching.
  4. You want to move to a distro with a stable upgrade path. Solution: Switch to FreeBSD if possible or use a less buggy distro like Oracle Enterprise Linux (though it still has some nasty landmines during upgrades we've seen, it's better than RHEL's terrible “backup everything and hope real hard.” upgrade procedure).

Benefits of Migration

At this point, it's very difficult to recommend Red Hat's products. There are a few exceptions where they have little or no competition, but mostly those are layered application products. So, customers who migrate will get five big benefits.

  1. Smaller costs and lower cost growth for non-production environments.
  2. More stability and some insulation from Red Hat's historically poor decisions.
  3. Direct and permanent open Internet access to your basic package repos without hassle-servers (Uh, I mean caching servers ala Satellite & Spacewalk).
  4. More stable upgrade paths between major versions
  5. Access to advanced storage technology Red Hat doesn't yet support.
moving_from_rhel_to_centos_or_oracle_linux.txt · Last modified: 2020/06/03 16:55 by sgriggs

Donate Powered by PHP Valid HTML5 Valid CSS Driven by DokuWiki