Gaia-8000 planning

Hey Cosmonauts,

We are beginning planning for gaia-8000.

Things to discuss in this thread are the validator ceremony for gaia-8000 and any other thoughts or questions.

A short summary of features for gaia-8000

Lots of UI changes
Internal code structure changes
Randomized testing
Many bug fixes (mostly minor)
Add transient store
Msgs are no longer run on CheckTx

Gaia-7003 spun up pretty nicely. Assigning more steak to the more active and engaged validators who are likely to come online promptly, while still having a larger set of smaller/newer/still learning validators participating is a good compromise.

This is only a guess, but seems like the distribution of bonded atoms at mainnet genesis will probably be similar to this pattern.

First of all I want to thank the Cosmos team for their great work and effort to continuously improve the network. It is very exciting to see not only the networks stability and capability grow, but also the community. As an active part of the community and a validator on the testnets for quite some time, I would however like to discuss some things about the ceremony and the genesis file. We all know that the decision for gaia-7003 happened quite spontaneous. I was online most of the night from Thursday to Friday to attend the ceremony of gaia-7002, the finding of the bug and how we proceed afterwards. As we all know, not everyone was online for the ceremony since we all live in different time zones and people need to sleep sometimes (yeah aurel I know, validators never sleep ). I think we can improve participation in the ceremony by better coordinating and communicating the time it happens. Therefore, I would like to propose that we vote on the time of the next ceremony, so that we find the time where most people can be online. Also I would like to know how “our most active contributors” were selected for increased stake in the new genesis file. While I understand the need to…

Excerpt (1198 of 1981 characters). Read the whole post on the forum ↗

These are great questions.

The configuration of the genesis state in gaia-7003 was made mostly my being tired and wanting to get a genesis block out Wednesday night.

We can fix the steak distribution by having the faucet send out 1000 steak to every validator currently online. That would even out distribution. I’ll synch with @Greg on this on Monday if we are still up.

@jlandrews made the argument that we didn’t really have enough data from 7002 about who was really online which made sense.

Yeah I totally agree we should even things out a bit more for gaia-8000.

With the release of Gaia8000, are we going to take the current state, or have a new registration?

I would prefer not to do another registration and instead state with current state with some modifications.

If there is a new registration, I believe there will be a lot of never online validators again. It in turn punishes those online validators and those who want participant in the testnet.

Just an idea: How about we start with a solid group of validators that we know will keep the testnet running and give them each quite a large share of steak. New validators can receive 5 steak (a small amount) from the faucet and prove that they want to actively participate in the network. If they stay online for one week (just a suggested period of time), they can start a proposal on the testnet. If accepted, each large validator sends them a share of their steak so that they get a major validator themself.

I know this would be a lot of work, but it would give everyone the chance to prove himself and “never online validators” would only account for a small share of steak. Also, we would have a great use case to try out voting on proposals. The issue with this is that we would rely on the big validators following the rules and we would see a certain degree of centralization.

My suggestion today with 7004 was to just have iqlusion (or another team members node(s)) take a very large amount of steak, then distribute out portions to diversify the steak as nodes come online. The problems we had before with registrations was trying to outvote the steak that wasn’t online, and this would not only allow us to get up and running, but once the steak is spread out amongst actives we can function like a faucet for the team (as it seems faucet functionality is hit or miss at times).

I understand it kind of goes against the grain to centralize a large portion of the steak, but I think we have had enough time playing with 65-67% active steak that it isn’t the primary focus of bug testing at the moment, and bringing a testnet up, live, and running as soon as possible provides a better environment for the whole in a more efficient manner.

i thank u for returning the 50steak that i lost due to unbonding bug & 20 steak out of nowhere. im quite pissed that proceau/delegate-now shows up on 7004 launch & gets 1000+ steak. better give weight to long time helpful testers adrian, aurel, chris remus, greg, jack, katernoir, kwun yeun, melea, pooch, ryan figment, suyuu, etc.

So what Tendermint is currently planning for is gaia v0.24 release between Friday Aug-3rd and Tuesday Aug 7th.

Depending on when v0.24 is released with can expect a gaia-8000 genesis to be published between Monday Aug 6th and Thursday August 9th.

If gaia-7004 crashes in the meantime, we will almost certainly just start a gaia-7005

I’m pretty sure that - if gaia-7004 crashes - we will have gaia-7005 up in no time. Validators are now used to the process and act faster.

Regarding gaia-8000, since @zaki gave an estimated timeline, we could now discuss how to find the best time for a ceremony with as many validators online as possible.

Does 8000 still imply the start of Game of Steaks? If so, I have some thoughts about launching the two in coordination w/each other. If not, I’ll save the comments for now.

We need fees implemented to do game of steaks.

Issue: Implement Piggy Bank Fee-Distribution Module

First update the spec then implement

If we don’t have fees implemented, I can’t use voting power as a proxy for success.

For v0.24

We are waiting on a set of PRs to be merged.

Keep an eye on the PRs below to estimate when we might start the gaia-8000 engines.

R4R: Staking Gaia-lite (ex LCD) refactor

R4R: Fix Cliff Validator Update Bugs

RRR4R: rational -> decimal

Tracking issue for Gaia v0.24 is up https://github.com/cosmos/cosmos-sdk/issues/1935

Quick update on the plans for gaia-8000

We have a tagged release candidate and are running internal testnets. We still need to fix a bug in gaiad export and run the automated fuzzing tests for some time.

I think we could realistically have gaia-8000 on Friday or Monday. Validators what do you prefer?

I’m comfortable with both options, but find it hard to turn down a weekend off

the day than is ready, is ok for me.

zaki:

I think we could realistically have gaia-8000 on Friday or Monday. Validators what do you prefer?

I prefer on Friday if gaiad-8000 is ready.

Friday looks good to me

Msgs are no longer run on CheckTx

I wonder that why we no longer run msgs on CheckTx. If we do that, how can we know is the tx valid.

I would much prefer Monday before Friday. Otherwise I won’t be able to participate.

Monday gets my vote.

I prefer Monday than Friday.

I prefer Friday if ready by then.

We prefer Monday over Friday!

BkHwlGBIgAEsSP7.jpeg600×800 41.1 KB

Just stating a general rule that helps sleep better

Monday, if possible

Okay update. Leaning towards Monday. We need to sort out a bug that caught in the simulator.

Issue: Panic in staking caught by randomized simulation

On the latest version (0.24.0-rc1 just merged in) of #1783:
--- FAIL: TestFullGaiaSimulation (206.41s)
panic: cliff validator has not been changed, yet we...

Those of you saying “Monday not Friday”, please post your node name/profile moniker so we know which validators to avoid upon launching of mainnet.

I’m sorry but it was a little unclear for me, whether the update has been pushed to Monday or not.
Are we still doing Friday, or are we going for Monday as of now?

yutian:

I wonder that why we no longer run msgs on CheckTx. If we do that, how can we know is the tx valid.

We still validate the transaction and check signatures and fees. We just don’t run all the additional state machine logic. Similar to ethereum - in the mempool they just check fees/amount, but don’t run the virtual machine until the tx is included in a block.

Current status:

Checking in with the SDK team about launching gaia-8000 when they cut v0.24 without an internal testnet for 24 hours.