[release/1.3.1] [DEVOPS-398] Testnet config

[rvl]

Description

This PR contains:

• Part of testnet configuration, cherry-picked from [DEVOPS-398] Script to launch demo cluster with new testnet genesis #3091 and [DEVOPS-398] Update testnet launch genesis data #3243 (develop branch).
• Docs with an explanation of testnet parameters.
• A fix for CBR-414, cherry-picked from [CBR-414] fix demo-nix.sh to have all core nodes working #3582, so that "testnet" and "mainnet" demo clusters can both be launched for testing.

Linked issues

https://iohk.myjetbrains.com/youtrack/issue/DEVOPS-398
https://iohk.myjetbrains.com/youtrack/issue/CO-354
https://iohk.myjetbrains.com/youtrack/issue/CBR-414

QA Steps

nix-build -A demoClusterLaunchGenesis -o demo-cluster-launch-genesis.sh
./demo-cluster-launch-genesis.sh

[mhuesch]

I'm confused by one part, which I commented on with a "❓"

[mhuesch]

👍 (assuming that "dev core" means this net will emulate mainnet/staging in development)

[rvl]

Yes I would like any dev clusters to work more like mainnet than testnet.

[mhuesch]

👍 mainnet has NMMustBeNothing

[mhuesch]

❓ I'm not sure why this is NMMustBeJust when the above mainnet config was NMMustBeNothing (but I'm also not very good and reading these config files)

[intricate]

Also wondering about this..

[rvl]

Thanks for ❓. This section describes the testnet genesis parameters. The << mainnet_... means to "inherit" config from the section labelled &mainnet_.... So my intention is to set up the testnet genesis data with most of the same parameters as mainnet, except overriding things such as protocolMagic.

[mhuesch]

Ah ok, that makes sense. Thanks for clarifying.

[mhuesch]

👍 This is consistent with https://github.com/input-output-hk/cardano-sl/pull/3585/files#diff-61feea0f7a7bc722b74eedc3907be256R257

[dcoutts]

Looks correct, but...

I worry order of overriding here however, that it could lead to mistakes too easily. Here we have a mainnet setting for Nothing, and then the testnet inheriting mainnet and overriding it back to Just. This seems fragile. I'd prefer if we can arrange things so that we only ever override from the default of Just to the special value of Nothing, and do so only for the "leaf" configurations, not for anything else that gets inherited later.

Could we do that? Have nothing inherit from the mainnet config itself, so that the mainnet config is a leaf config. So for example instead of of putting the Nothing into mainnet_base, where it gets inherited by the testnets, put it into the mainnet config only (which I assume it a leaf config). Then the testnets would not need any special override.

Summary: can we follow the rule that we only use requiresNetworkMagic: NMMustBeNothing and not NMMustBeJust, and we only use it at "leaf" configurations, not configurations that are inherited from. I think this rule will make it much more explicit and harder to make silly mistakes.

[rvl]

Thanks @dcoutts . I have reset the mainnet_base genesis spec protocolMagic, so that there is no double-overriding of the setting. In addition, I made all of the testnet_launch genesis spec protocolConstants explicit and not inherited just to try and improve clarity.

[jmitchell]

We need to ensure any legacy clusters, like staging and mainnet, explicitly pass requiresNetworkMagic: NMMustBeNothing. Otherwise, they will default to NMMustBeJust, which is the new address behavior for testnet.

cc @mhuesch

I have implemented your suggestions in #3659