Sui’s three outages expose ‘blast radius’ risk – Is the 15% drop in price a start?

On the 28th and 29th of May, there were three outages at the Sui Foundation.

Due to these network stalls, transaction processing and settlement were momentarily disrupted, which caused significant chaos in the Sui [SUI] ecosystem.

After realizing the risk involved in the current proposal, the Sui Foundation came up with a strong solution

The very first network stall

A bug in Sui’s new address balance and hybrid gas payment system was the reason for the first outage.

Rarely, two transactions attempting to spend the same money at the same time would result in one being correctly canceled with an InsufficientFundsForWithdraw error.

However, in this case, the canceled transaction attempted to spend money and went through gas smashing, resulting in an invalid negative balance at settlement

Sui noted,

Cancelling transactions with this error is how the scheduler prevents overdrafts, but it cannot do this if the canceled transaction still debits funds due to gas smashing.

Nonetheless, the Sui developers were able to fix this bug by preventing gas smashing after a transaction has been canceled for lack of funds.

The second outage

The network recovered rapidly after the initial fix stopped gas smashing for transactions canceled because of InsufficientFundsForWithdraw.

An edge case, however, made it possible for the insufficient-funds error to be concealed by another cancellation reason because transactions can fail for various reasons.

Bypassing the protection, this resulted in a second outage and the same balance underflow bug. After resolving these overlapping failure scenarios, developers implemented a more thorough patch that stabilized the network.

The third mainnet halt

Fortunately, the third outage was not brought on by gas payments. Instead, it was caused by a bug in Sui’s Distributed Key Generation (DKG) system.

DKG failed because there was not enough participation after validator restarts, but the failure status was not recorded.

Because of this, randomness-dependent transactions were stuck, and the network was unable to complete its planned epoch transition. This happened because validators had restarted without realizing DKG had already failed.

To resolve the problem, developers added a mechanism to safely close a stuck epoch and made sure DKG status remained consistent across restarts.

Interestingly, Sui concluded it best when they noted,

Today, the system lacks a defense-in-depth layer that would bound the blast radius of such a crash.

Sui market dynamics

This coincided with the SUI price dropping from $0.998 to $0.8783 at press time, accounting for over a 15% drop in the past week.

Meanwhile, the Open Interest of Sui suggested that the traders were adding to their positions instead of closing them. This further confirmed that sellers were regaining strength against the buyers.

Source: CoinGlass

Final Summary

• In less than 48 hours, three different bugs caused three outages, revealing flaws in Sui’s ecosystem.
• Sui’s outages were the result of a series of uncommon edge cases rather than a single failure.