After the somewhat imperfect launch of the Medalla Eth 2.0 testnet, developers discuss what initially went wrong.
Following the successful but imperfect launch of the Medalla testnet, the first community testnet for Ethereum 2.0, developers spoke on the issues and needed improvements highlighted by the event.
In an Eth 2.0 Dev call held on Aug. 6, the network’s client developers discussed what initially went wrong.
As Cointelegraph previously reported, it took some time for the Blockchain to achieve finality, with only 57% of the stake participating in consensus. Generally, 70% is the minimum for a network to be healthy, as outlined in a community explainer. Participation is currently sitting at a stable 80%, according to beaconcha.in.
There were several issues outlined by the team that contributed to this slow start.
Before the launch, client developers discovered several “peering” bugs that would have prevented effective connections between nodes. As they rushed to fix these just a day or two before block genesis, some validators may have failed to update to the latest versions.
They also identified several issues with the launchpad used to onboard users interested in staking their coins. Some of these related to user experience and education, as the teams were unable to prepare node binaries in time. Furthermore, some users were found to have made multiple 32 ETH deposits, likely with the misguided expectation that these would let them have more validators.
Manually setting up multiple validators also became an issue due to MetaMask, which opened new popups for each 32 ETH batch all at the same time.
Offline stakers and technical issues
During launch, stake participation was just 57% during the first hour. This was, according to the team, partially due to Nimbus and Lodestar clients having low participation in attestations, which is what verifies each proposed block. This accounted for about 10% of the gap.
The remainder, in addition to the confusion with last-minute client updates, was due to some stakers being offline.
This is to be expected, they said, and the network will “leak” the 32 ETH stake until the offline validators are kicked off the network. The incentive to not lose money should be enough of a deterrent, but since testnet Ether has no value, this crypto-economic incentive did not fully work.
Some of the large stakers did not spin up their clients. Developers contacted one of them, and through a combination of factors, participation was quickly brought to about 80%.
The network is nevertheless not completely stable. Prysm and Lighthouse, the two leading clients according to popular usage, are still seeing validation issues.
The team noted that the dominance of Prysm and Lighthouse should be dealt with. According to polls cited by Ethereum Foundation’s Danny Ryan, about 90% of all nodes run those two clients.
Most teams focused on bug fixing this week. Some bugs are allegedly more severe than others, but none of the clients are yet perfect.
Research and development continues as Vitalik Buterin published an annotated specification for Ethereum 2.0 Phase 1, which should include basic sharding.