You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
During Testnet 72, we tried to ship a hotfix for a DEX problem (#4239), which failed catastrophically due to tombstoning of the PL testnet validators. As part of preparing for that upgrade, we decided to submit an emergency-halt governance proposal, as an exercise of remediation procedures suitable for mainnet.
While the emergency-halt did properly stop the chain as soon as sufficient stake weight voted for it, the threshold was too high for the functionality to be useful: 1/3 of validators (+1) could simply shut down, resulting in the chain halting. We should lower the threshold for an emergency-halt vote to pass, specifically to 1/3, so that validators can effect an organized halt with the same quorum that already can, if coordinating out of band.
Notably if 1/3 + 1 validators coordinator out of band to shut down manually, that'll still stop the chain, but recovery would be arduous, as state surgery will be required to reconstruct an agreed upon resumption point. Using a governance mechanism prevents this confusion, ensuring all validators and nodes have the same picture of the network at time of halt.
The text was updated successfully, but these errors were encountered:
This changes the ratio in tally.rs emergency_pass() to 1/3 and also
updates any comments/docs referring to the previous 2/3 threshold. Note:
this will also affect `IBCFreeze` and `IBCUnfreeze`, so those have also
gotten easier to accomplish (seems fine for the same reasons, but wasn't
explicitly noted in the ticket). Closes#4389.
During Testnet 72, we tried to ship a hotfix for a DEX problem (#4239), which failed catastrophically due to tombstoning of the PL testnet validators. As part of preparing for that upgrade, we decided to submit an emergency-halt governance proposal, as an exercise of remediation procedures suitable for mainnet.
While the emergency-halt did properly stop the chain as soon as sufficient stake weight voted for it, the threshold was too high for the functionality to be useful: 1/3 of validators (+1) could simply shut down, resulting in the chain halting. We should lower the threshold for an emergency-halt vote to pass, specifically to 1/3, so that validators can effect an organized halt with the same quorum that already can, if coordinating out of band.
Notably if 1/3 + 1 validators coordinator out of band to shut down manually, that'll still stop the chain, but recovery would be arduous, as state surgery will be required to reconstruct an agreed upon resumption point. Using a governance mechanism prevents this confusion, ensuring all validators and nodes have the same picture of the network at time of halt.
The text was updated successfully, but these errors were encountered: