We opened this discussion before (Propose to increase Block Gas Limit to 35.000.000 gas) to increase the gas block limit to 35M gas and after some feedback and more testing, we reach to a new proposal.
- Increase the gas block limit from 20M to 50M gas
- Decrease the gasPriceMinimum target density from 0.8 to 0.7
The motivation for the proposal remains the same (increase network TPS and prepare for higher demand). Which is also possible to address because of the same improvements we listed in the previous proposal.
As the testing show us that the network can handle a load of 50M gas, but as the nature of our tests were not completely equal to a production environment, clabs in the last proposal decided as a safe guard to reduce that maximum gas limit. This proposal changes that, but instead of reducing the gas limit, we are proposing to decrease the target density, which would discourage a sustained full block stress of the network for longer periods than 10 minutes (for a more detailed explanation check the proposal: CGP )
This would allow the network to absorb peaks of 50M gas without issues, and a sustained load of 35M gas without changes in the
We would love to hear your feedback on this revamp of the last proposal
Yeah this will be very nice, especially given that DEFI actions on UniV3 will definitely start eating more gas (a lot more for people who are LPing)
We also had pretty bad congestion during peak liquidation blocks over the last 72 hours as well.
As for validators who want to make sure their hardware is up for this, a few simple recommendations:
- if hosting on GCP, switch to TAU instances
- if hosting on AWS, switch to M6a instances
- if hosting on Azure, switch to Dasv5 or Dadsv5 instances
- if hosting on Hetzner, switch to EX43 instances
Any other platforms, run a quick geekbench5 benchmark and make sure that the single threaded performance is > 1200, the higher single threaded performance, the better. Do not use a machine type that scores < 1000 on the single threaded perf.
There were a few blocks today when we ran out of block space to run all of the eligible moola liquidations.
I also suspect we had brief periods when Oracle updates at 0.6gwei was being priced out of the block. Does the oracle updater code run with a fixed gas 0.6gwei or does it have a dynamic mechanism to run with higher gas price during congestion?
Yes, the oracles have a dynamic way to calculate the gas price, being a lot bigger than the usual ones.
Also, as far as I know, some validators have the oracles whitelisted, which implies that those txs are proposed as locals, which will have a bigger priority