Gas Reference
Use this page as the condensed cheat-sheet for gas-related settings and formulas on Sei. Every entry references the release where the behaviour was introduced so you can map it to your node fleet.
Quick Lookup
max_log_no_block | 10000 logs when no range is supplied. Increase only if you batch replay historical events. |
max_blocks_for_log | 2000 block span per request. Larger values multiply DB pressure; shard queries instead. |
max_subscriptions_new_head | 5000 concurrent eth_subscribe:newHeads feeds. Lower for constrained hardware. |
max_concurrent_trace_calls | 10 concurrent tracer calls. Raise cautiously and pair with rpc_trace_pending alerts. |
trace_timeout | 30s per trace. Tighten if custom tracers risk monopolising threads. |
Gas Formulas
| Transaction receipt | gasUsed = totalGasConsumed - refund. Refund is capped at 20% of gasLimit in line with go-ethereum rules. |
| Block total | block.gasUsed = sum(receipt.gasUsed). Accurate for blocks produced after v6.1.11. Earlier heights may diverge. |
| Estimation buffer | suggestedGas = estimate * 1.05. Apply a 5% safety margin for user-facing UIs. |
Release Notes
- v6.1.11
- Block
gasUsedaligned with Tendermint execution (sei-chain@1efdec1eb). - Pointer and Solo precompile flows warmed in
eth_estimateGas. - Synthetic logs tagged with
synthetic: true.
- Block
- v1.15.7-sei-7
- Tracer length guard prevents malformed frames from panicking.
- v1.15.7-sei-6
- RPC now returns error payload with trace data when custom tracers panic.
Operational Tips
- Keep
max_blocks_for_logconservative on shared endpoints; let power users run their own archive node if they need large spans. - After upgrades, clear any CDN or proxy caches so clients immediately pick up corrected
gasUsedvalues. - Use Grafana or Prometheus exporters to monitor
rpc_gas_usedandrpc_trace_pendingmetrics if you run the RPC server fromsei-chain.
⚠️
Always restart both the Tendermint process and the embedded RPC server once you upgrade to a release that changes gas accounting. Partial restarts can keep stale caches alive.
See Also
rpc-gas-accounting– Full explanation of execution flow and validation steps.rpc-regression-playbook– Suggested QA coverage for RPC consumers.tracing-playbook– Panic handling for tracers and debug sessions.
Last updated on