Context & versions
Full blocks sometimes take over 2 seconds to produce, which leads to forks. This is more apparent on longer chains.
Additional context from Tomasz Bartos: "due substrate receives incorrect weight, I was writing posts about it in the node/perf channels few times. For instance we do not count validation time, only the execution so that is the first and significant mismatch."
Related Slack discussions:
While this can be controlled with pool-size, it's not a perfect solution. The ledger weights should be revisited and targeted for a fix soon after launch.
Steps to reproduce
- Run the chain with high transaction throughput to produce full blocks
- Observe block production times exceeding 2 seconds
- Observe resulting forks, more apparent on longer chains
Actual behavior
Full blocks take over 2 seconds to produce because substrate receives incorrect weight (validation time is not counted, only execution time). This mismatch causes forks.
Expected behavior
Block production should stay within the 2-second slot time even with full blocks. Ledger weights should accurately reflect both validation and execution time.
Migrated from PM-22396
Context & versions
Full blocks sometimes take over 2 seconds to produce, which leads to forks. This is more apparent on longer chains.
Additional context from Tomasz Bartos: "due substrate receives incorrect weight, I was writing posts about it in the node/perf channels few times. For instance we do not count validation time, only the execution so that is the first and significant mismatch."
Related Slack discussions:
While this can be controlled with pool-size, it's not a perfect solution. The ledger weights should be revisited and targeted for a fix soon after launch.
Steps to reproduce
Actual behavior
Full blocks take over 2 seconds to produce because substrate receives incorrect weight (validation time is not counted, only execution time). This mismatch causes forks.
Expected behavior
Block production should stay within the 2-second slot time even with full blocks. Ledger weights should accurately reflect both validation and execution time.
Migrated from PM-22396