Why is the Rocky GA kernel based on errata kernel instead of RHEL GA kernel? IIRC this started with Rocky Linux 9.5. After seeing the staging builds of Rocky 9.6 RC1 and 10.0 RC1, it appears to be a similar pattern.
For example:
It is simply easier to build and release a newer kernel because we cannot release the same day as our upstream, it’s simply not possible. Doing this also reduces time it takes to get secure boot builds done in a reasonable time frame and allows our users to have a close to up-to-date kernel as possible. We purposely post on our wiki what kernel we start our releases with to make it clear what we start with, just so everyone is aware.
I know you are mentioning the kernel, but there are many other packages that end up being newer than what was released by upstream that are not the kernel. We generally don’t see any issues with the way that we release Rocky Linux to our users, so it would help to understand why this may be considered an issue.
Thanks for the quick reply. I’m an ELRepo dev, so I notice the kernel version differences.
So far it hasn’t been an issue (knock on wood). But there is the potential that if the RHEL kABI gets broken in the errata kernel that Rocky uses for GA, then it could break compatibility with some of our kmod DUD’s…specifically for bootable storage devices.
That is a very good point and something I’ll keep in mind for future releases. It is true that ABI breakage can happen even in the middle of a “stable” release. We’ve seen it happen before too.
Since we are transitioning our build systems around, it may be easier for us to match the release kernel. I know historically in the past we were able to, but given the weird place our build infra is in, some things are a bit more difficult than they should be. Either way I’ll definitely keep this in mind and let my team know as well.