[BlockSTM] Per-block Gas Limit

[danielxiangzl]

Description

This PR supports early halting parallel execution (BlockSTM), for scenarios such as exceeding per-block gas limit, module publishing read / write conflicts, reconfiguration and VM internal error. For per-block gas limit, a new on-chain config is added for compatibility.

The PR builds upon rolling commit, which allows the BlockSTM to keep track of committed transaction prefix, and therefore compute the accumulated gas of committed transactions.

Once the parallel execution needs to be early halted, one thread will be in charge of waking up any pending thread (on dependency reads) and mark the execution status of each transaction properly (a new status named ExecutionHalted is introduced).

Fixed related tests on parallel executor and executor. Added on-chain config for compatibility. Passed forge test with block gas limits. Additional monitoring metrics are added in the Execution dashboard with row named "Execution Per Block Gas".

TODO:

• Determine the per-block gas limit.
• Properly remove the storage error logging in MoveVM, as we return the storage error for read dependency once the parallel executor is early halted. The error should be passed to AptosVM and handled by the speculative logging.

Test Plan

Existing and new unit / prop tests on parallel executor and executor with block gas limit on and off.
Smoke test related to execution / storage.
Forge performance test and forge compatibility test.
Will add additional tests if necessary.

[gelash]

Current semantics seems that we skip the rest of the block once accumulated gas exceeds the limit, but include the first transaction that exceeded (made it >= to be precise). Is this what we want, or should we include the prefix of txns such that the total consumed gas is <limit? (or <=, but I guess < is fine here).

@danielxiangzl @zekun000 @grao1991

[gelash]

could merge this check with must_skip || accumulated_gas >= PER_BLOCK_GAS_LIMIT?

[danielxiangzl]

Actually need to separate them to distinguish reconfiguration from exceeding the per-block gas limit.

[davidiw]

Do we have a (public) design doc on this topic?

[zekun000]

ugly

[github-actions]

✅ Forge suite compat success on testnet_2d8b1b57553d869190f61df1aaf7f31a8fc19a7b ==> dfd15ce3648718b673b745b133ad21757cbf6daf

Compatibility test results for testnet_2d8b1b57553d869190f61df1aaf7f31a8fc19a7b ==> dfd15ce3648718b673b745b133ad21757cbf6daf (PR)
1. Check liveness of validators at old version: testnet_2d8b1b57553d869190f61df1aaf7f31a8fc19a7b
compatibility::simple-validator-upgrade::liveness-check : 10187 TPS, 3682 ms latency, 5800 ms p99 latency,no expired txns
2. Upgrading first Validator to new version: dfd15ce3648718b673b745b133ad21757cbf6daf
compatibility::simple-validator-upgrade::single-validator-upgrade : 5526 TPS, 7197 ms latency, 10000 ms p99 latency,no expired txns
3. Upgrading rest of first batch to new version: dfd15ce3648718b673b745b133ad21757cbf6daf
compatibility::simple-validator-upgrade::half-validator-upgrade : 5771 TPS, 6865 ms latency, 9000 ms p99 latency,no expired txns
4. upgrading second batch to new version: dfd15ce3648718b673b745b133ad21757cbf6daf
compatibility::simple-validator-upgrade::rest-validator-upgrade : 8056 TPS, 4685 ms latency, 8800 ms p99 latency,no expired txns
5. check swarm health
Compatibility test for testnet_2d8b1b57553d869190f61df1aaf7f31a8fc19a7b ==> dfd15ce3648718b673b745b133ad21757cbf6daf passed
Test Ok

• Grafana dashboard
• Humio Logs
• Pyroscope Profiling
• Test runner output
• Test run is land-blocking

[github-actions]

✅ Forge suite land_blocking success on dfd15ce3648718b673b745b133ad21757cbf6daf

performance benchmark : 6060 TPS, 6522 ms latency, 26800 ms p99 latency,no expired txns
Test Ok

• Grafana dashboard
• Humio Logs
• Pyroscope Profiling
• Test runner output
• Test run is land-blocking

[github-actions]

✅ Forge suite framework_upgrade success on aptos-node-v1.3.0_3fc3d42b6cfe27460004f9a0326451bcda840a60 ==> dfd15ce3648718b673b745b133ad21757cbf6daf

Compatibility test results for aptos-node-v1.3.0_3fc3d42b6cfe27460004f9a0326451bcda840a60 ==> dfd15ce3648718b673b745b133ad21757cbf6daf (PR)
Upgrade the nodes to version: dfd15ce3648718b673b745b133ad21757cbf6daf
framework_upgrade::framework-upgrade::full-framework-upgrade : 6554 TPS, 6035 ms latency, 9600 ms p99 latency,no expired txns
5. check swarm health
Compatibility test for aptos-node-v1.3.0_3fc3d42b6cfe27460004f9a0326451bcda840a60 ==> dfd15ce3648718b673b745b133ad21757cbf6daf passed
Test Ok

• Grafana dashboard
• Humio Logs
• Pyroscope Profiling
• Test runner output
• Test run is land-blocking