doc: add a page explaining storage class parameters

[james-munson]

longhorn/longhorn#4776

[shuo-wu]

1. It's better to add one example at the beginning.
2. I think we can explain how Longhorn creates the default SC. And why we use a ConfigMap to create that default SC (since some fields are immutable.)

[shuo-wu]

In general LGTM

[ejweber]

Much improved. A couple of questions still. One big thing we need to do is "backport" this to other versions of the website. Probably we don't need to go all the way back to the beginning of time, but to the versions that aren't EOL. That will be annoying, as some of these options are new.

[innobead]

In general, LGTM. Just some minor suggestions.

Good job, James.

[PhanLe1010]

The PR LGTM once all the above review comments are resolved.

[ejweber]

It looks like you addressed every comment of mine but one. This is looking really good. Still need to add a version of this (minus newly introduced parameters) to other versions IMO. Let me know if you disagree.

[innobead]

LGTM. The only remaining from my side is to have this page for the latest existing patch or upcoming patch release doc as well, such as 1.3.3. 1.4.3 & 1.4.4, 1.5.1 & 1.5.2.

P.S. No need to have that in 1.3.4, since it's EOL and we will not release it.

[james-munson]

I added the 1.4.x and 1.5.x versions.
1.4.x are all identical.
1.5.x are all identical to each other, and differ from 1.4.x in that the backendStoreDriver (v2 engine) parameter is added.
1.6.0 differs from 1.5.x in that the replica disk soft anti-affinity is added. (Looking ahead to when Eric's PR is merged! It will need one tiny tweak when that happens, to point to the global setting, which doesn't exist yet, and gives an invalid reference in the build.)

I did not add a 1.3.x version. If we do that, it will also require adjusting filesystem trim and mkfs parameters. WDYT? Worth going back that far?

[james-munson]

Also added documentation for "replica disk soft anti-affinity" to global settings and replica scheduling section.

[ejweber]

Thanks for the Replica Disk Soft Anti Affinity addition. Looks good!

I am personally fine with not including this page on the v1.3.x branch given its EOL status, so approving. Not sure where others will fall.

[PhanLe1010]

I am fine with not having it in 1.3.x too

[innobead]

LGTM, but some comments need to resolve first.

[innobead]

The value should be v1 or v2. If it's empty, it's just v1.

/k8s/pkg/apis/longhorn/v1beta2/volume.go#L162-L165

const (
	BackendStoreDriverTypeV1 = BackendStoreDriverType("v1")
	BackendStoreDriverTypeV2 = BackendStoreDriverType("v2")
)

[james-munson]

Good catch. Fixed, thanks.

[innobead]

Why is it not false by default? A disk with two replicas of a volume seems not a good practice.

BTW, Node Soft Anti Affinity will supersede this setting if it's false. It's better to mention this.

[innobead]

oh, it was reviewed. It's about backward compatibility. https://github.com/longhorn/longhorn/blob/master/enhancements/20230718-disk-anti-affinity.md#upgrade-strategy.

[james-munson]

Added, thanks.