Automated Local Disk Setup

[zimnx]

The main painpoint of ScyllaDB on Kubernetes was configuring the
disks. Each cloud provider sets them up in various location, under
different names. Different instance types comes with different number of disks
and different sizes.
To improve user experience we introduce automatic local disk setup.

v1alpha1.NodeConfig API was extended with additional setting related to
local disks setup.
Scylla Operator will manage a DaemonSet responsible for creating an
RAID0 array out of discovered disks, formatting it to desired filesystem and
mounting in provided location together with provided mount options.

This feature is considered experimental, and is subject for deprecation
and incompatible changes. Using it, may require manual steps during
upgrades.

Fixes #1096

Requirements:

• Bump Kubernetes version to 1.24.9 in CI #1106

[tnozicka]

part 1

[tnozicka]

https://pkg.go.dev/golang.org/x/exp/constraints#Signed

[zimnx]

https://pkg.go.dev/golang.org/x/exp#section-readme

Warning: Packages here are experimental and unreliable. Some may one day be promoted to the main repository or other subrepository, or they may be modified arbitrarily or even disappear altogether.

[tnozicka]

I though it might not have been there when you started.

What's the worst case to happen when they remove it? I assume we'd "fork"/copy that file and maintain it on our own. But if they keep it, we don't have to review, maintain and update our copy, so to me it seems like we can only gain or end up in the same place. They will likely just move it so we only endup adjusting the import path at some point. And we have version pinning and vendoring.

btw. we already use a lot of deps in v0.x.x that have the same (non)commitment.

[tnozicka]

There is a lot of good work in this PR! And a lot of pain felt when trying to figure out all the compatibility with different mdadm versions / setups.

Modulo the regex matching I am ready to land this when the CI gets green.

Consistency wise, please make sure this returns 0

git diff HEAD..master -- ./cmd/ ./pkg/ | grep -i -E -e 'if .*;.*{' | wc -l

[zimnx]

Consistency wise, please make sure this returns 0

git diff HEAD..master -- ./cmd/ ./pkg/ | grep -i -E -e 'if .*;.*{' | wc -l

These aren't inconsistent

$ grep -rni --exclude-dir=vendor -E -e 'if .*;.*{' | wc -l
3692

nor I don't get why this syntax suddenly started to be pointed out without any argument or discussion behind.
Although, I changed all added ones to suggested syntax - we are circling around pointless changes for too long, lets start moving forward.

[tnozicka]

nor I don't get why this syntax suddenly started to be pointed out without any argument or discussion behind.

suddenly is a bit strong word - I recall the 2 of us having a discussion about it in the past and the readability and length line was pointed out on several PRs before. Given the magnitude of this PR I felt the grep command is easier then individual comments. If you disagree, I imagine there are more polite ways to raise it, same way as the other point you were trying to make.

Although, I changed all added ones to suggested syntax - we are circling around pointless changes for too long, lets start moving forward.

This is not something that's holding this PR.

[tnozicka]

thanks for the updates and the enhanced regexp logic. the only major one open I see is to use the fixed image in #1107 (comment)

[tnozicka]

/approve
/lgtm