Ethereum - A week after Shapella
The Shapella upgrade, which happened on April 12, 2023, has been the most significant event for Ethereum since The Merge in September 2022. In our recent article, we explored the important features of the Shapella upgrade, with a focus on partial and full withdrawals. In this article, we will go deeper into the upgrade. We will look at the current network status a week after the upgrade, full and partial withdrawals, and who withdrew. We also examine how this transition affected the performance metrics of pools and operators.
Table of contents
Network Overview
One week after Shapella, the current total staked amount is 17,972,763 ETH, 54k lower than the maximum on April 12th. This stake represents approximately 15.65% of the total circulating supply of Ethereum, indicating strong investor confidence in the platform. The network's effectiveness has decreased to 91.46% because it experienced minor struggles after the upgrade.
The First hours after the upgrade
The most noticeable inaccuracies occurred in the first three hours after the activation slot. During this period, around 15% of blocks were missed, 7% of attestations did not occur, and approximately 30% of head votes were incorrect. Naturally, these inaccuracies led to a decrease in the share of consensus layer rewards. It should also be noted that the correlation between incorrect head votes and missed blocks is observed due to consensus rules - the lifetime of attestation of a valid head vote block is one slot, and if it is not submitted, then there was no block.
Going into detail, we have identified several reasons for the network indicators observed in the first hours after the update. Firstly, the reduced percentage of correct attestations suggests that up to 9% of validators were not updated, which directly affected block misses. Secondly, Prysm experienced problems with obtaining MEV-blocks. It was unable to produce blocks while connected to relays. Thirdly, Lighthouse was 100% CPU-loaded for about 2 hours, which led to missed attestations and late block proposing. Another reason, in our experience, is that the Teku client took a long time, about 15 seconds, to import blocks, causing lag on the network. These facts highlight the importance of client diversity in ensuring network stability.
However, the minor nature of the problems is confirmed by the fact that with each epoch, effectiveness slowly but constantly improved, and in a day, though not without emergency releases of CL clients, it reached the usual indicators.
Additionally, with the Capella update, validators who specified the "old" BLS address as withdrawal credentials were able to change it to the ETH1 address with the 0x01 prefix. This operation requires more RAM, bandwidth, and CPU power from the CL Node in each slot and had a small impact on the degradation of network performance metrics.
In the first 6 days, 226k validators changed their addresses, it’s about 70% of all validators with BLS withdrawal credentials. This means that the time needed for one iteration of the withdrawal clock has increased significantly, but has not reached its maximum.
Withdrawals
The Shapella upgrade introduces two types of withdrawals: full withdrawals (also known as exits) and partial withdrawals (staking reward collection). If you would like to dive deeper into how withdrawals work, you can explore our Shapella Upgrade article.
After the Ethereum update, the total amount of Ethereum withdrawn was found to be 1,323,637 ETH, with only 1.6% of validators exiting the system. This withdrawal led us to analyze the amount of Ethereum that was sold on DEX, and we found that only 0.25% of the total ETH was actually sold.
The analysis also revealed that most of the validators who withdrew were from Kraken, accounting for over 90% of the total withdrawals. The reason behind this mass withdrawal was a fine imposed by the SEC, which forced the exchange to wind down its US staking operations.
We can see that a large number of full withdrawals occurred between April 15th and April 20th. The Kraken validators' indexes are sequential, meaning their withdrawals also occurred sequentially. As a result, we observe large volumes on those days. All 333k ETH withdrawn from Kraken's validators is currently being held in the withdrawal address 0x210b3cb99fa1de0a64085fa80e18c22fe4722a1b.
In addition, it’s important to note that there are currently 19,029 validators waiting to exit the active validator set. Currently, only 8 validators can exit every epoch. After that, there is a waiting period of 27 hours to ensure that the validator is unslashed. Finally, the withdrawal process involves a looped queue that can take up to 5 days. Therefore, at this time, the exit queue is approximately 12-16 days.
The number of validators in the exit queue is gradually decreasing. Our analysis suggests that over 72% of active exiting validators belong to centralized exchanges (CEX). Meanwhile, the number of validators waiting to enter the active validator set is increasing and currently stands at 8,152.
This trend indicates a growing interest in Ethereum staking, as evidenced by the observable increase in the waiting time within the queue. Currently, the waiting time is equal to 1 day and 20 hours.
Operators performance
We examined how this transition affected on validator’s performance metrics divided by pools and operators. As we can observe, after the update on 13th April validator effectiveness dropped over 10 p.p for operators.
The effectiveness of a validator is determined by its block proposal and attestation rates, which are the measures of performing the validator’s duties like timely block proposals and attestations. Let's take a closer look at each of these indicators.
After a recent update, there were a significant number of missed blocks in the network. On April 13th, there were a total of 633 missed blocks, which is 342% higher than the number of missed blocks on April 12th. As shown in the plot, this had a negative impact on the block proposal rate and led to a drawdown.
The attestation rate is measured by three factors: participation rate, correctness, and inclusion delay.
We can see that the participation rate didn't change much after the update for most validators, except for RocketPool, whose uptime dropped by around 2.5 percentage points. However, the correctness plot shows that there were many incorrect head votes and target votes after the update. Additionally, the average inclusion distance between the attestation slots attributed and the actual slots the votes were included in, also increased.
However, during the week, the operator's performance improved, allowing them to recover to the values of the level before the update.
Slashings
This week was unfortunate for several validators. There were 11 slashings on the next day after Shapella by validators of RockLogic GmbH in Lido pool, but this is not related to the update, just the operator suffered from a bug with keystore in the Prysm client. This case, certainly, was analyzed in detail, accompanied by a post-mortem and has already been fixed.
The good side of this news is that since Lido has a money-back policy and treasury, none of the clients will definitely suffer losses.
Conclusion
In conclusion, the Shapella upgrade has had a significant impact on the Ethereum network. The transition was not without its challenges, with a significant number of missed blocks in the network and a decrease in effectiveness. However, the network has shown resilience and has been able to recover to pre-upgrade levels in terms of performance. The increase in the waiting time for validators to enter the active validator set indicates a growing interest in Ethereum staking, which bodes well for the future of the network. While there were some issues with the upgrade, the Ethereum community has shown its ability to adapt and overcome challenges, which is a positive sign for the future development of the platform.