Checklist for Doing a Public, Coordinated Schlesi Testnet

[rauljordan]

Hi all,

As this multiclient effort matures, it's important for teams to be on the same page regarding what is needed before we can do a public, coordinated testnet launch like the @prysmaticlabs' topaz testnet here. Here are some tentative milestones before a lighthouse+prysm public testnet announcement should occur:

• No skipping or minimal finality delays for N epochs
• Testnet resilience to recover under N epochs without finality
• Initial chain sync stress tested
• Stress testing N number of validators, where N in (16k, 50k, 100k)
• Slashable events occurring and successfully caught by slasher
• Major node panics/crashes resolved across both clients
• High chain participation > 97% as well as good profitability metrics for validators (less critical as can be improved over time)

Thoughts?

[prestonvanloon]

Please add validator activation and voluntary exits.

[q9f]

These points are looking sane. It still requires a lot of work for a long-term public multi-client testnet. Currently, Schlesi will probably either die or be reset every other week. Right now, it still requires a lot of patching and should mainly be used by client teams to nail down remaining issues.

Lighthouse is pretty good at supporting custom chain configurations, so it's easy to setup a couple of LH nodes with a Schlesi config. Unfortunately, there is no network config abstraction in Prysm that could be used currently. On my wish-list for Prysm would be either support for a custom network specification (yaml?) to override the default beaconChainConfig or a way to fully specify networks interally, i.e., having a --chain topaz and --chain schlesi flag. Could you add that to the list maybe?

Also, spec version is always a moving target and in case we would have to reset Schlesi or Topaz today, I would go for v0.11.2 instead of v0.12.0. This would also allow other clients that implement v0.11.x to catch up with these networks soon.

A realistic timeline for a public Schlesi launch would be >6 weeks at least, as I would carefully guess. I'm not sure what the timeline for v0.12.x looks like but I believe we will be still on v0.11.x in June. Correct me if I'm wrong.