feat(vm-runner): shadow protective reads using VM runner (#2017)

## What ❔

Adds a new component `vm_runner_protective_reads` that computes
protective reads independently and asynchronously from state keeper.

For now, the component does not actually save anything; instead, it
computes protective reads and compares them against what state keeper
has already written to the DB. So, in short, this is just the first
stepping stone aka a sanity check that the VM runner mechanism works as
intended.

## Why ❔

In the future, we want to be able to save protective reads
asynchronously thus saving time on L1 batch sealing.

## Checklist

<!-- Check your PR fulfills the following items. -->
<!-- For draft PRs check the boxes as you complete them. -->

- [x] PR title corresponds to the body of PR (we generate changelog
entries from PRs).
- [ ] Tests for the changes have been added / updated.
- [x] Documentation comments have been added / updated.
- [x] Code has been formatted via `zk fmt` and `zk lint`.
- [x] Spellcheck has been run via `zk spellcheck`.

.github/workflows/ci-core-reusable.yml

@@ -104,7 +104,7 @@ jobs:
       # `sleep 60` because we need to wait until server added all the tokens
       - name: Run server
         run: |
-          ci_run zk server --uring --components api,tree,eth,state_keeper,housekeeper,commitment_generator &>server.log &
+          ci_run zk server --uring --components api,tree,eth,state_keeper,housekeeper,commitment_generator,vm_runner_protective_reads &>server.log &
           ci_run sleep 60
 
       - name: Deploy legacy era contracts
@@ -134,7 +134,7 @@ jobs:
         base_token: ["Eth", "Custom"]
         deployment_mode: ["Rollup", "Validium"]
     env:
-      SERVER_COMPONENTS: "api,tree,eth,state_keeper,housekeeper,commitment_generator${{ matrix.consensus && ',consensus' || '' }}"
+      SERVER_COMPONENTS: "api,tree,eth,state_keeper,housekeeper,commitment_generator,vm_runner_protective_reads${{ matrix.consensus && ',consensus' || '' }}"
 
     runs-on: [matterlabs-ci-runner]
     steps:
@@ -302,7 +302,7 @@ jobs:
     runs-on: [matterlabs-ci-runner]
 
     env:
-      SERVER_COMPONENTS: "api,tree,eth,state_keeper,housekeeper,commitment_generator${{ matrix.consensus && ',consensus' || '' }}"
+      SERVER_COMPONENTS: "api,tree,eth,state_keeper,housekeeper,commitment_generator,vm_runner_protective_reads${{ matrix.consensus && ',consensus' || '' }}"
       EXT_NODE_FLAGS: "${{ matrix.consensus && '-- --enable-consensus' || '' }}"
 
     steps:

core/bin/zksync_server/src/main.rs

@@ -13,7 +13,8 @@ use zksync_config::{
         house_keeper::HouseKeeperConfig,
         ContractsConfig, DatabaseSecrets, FriProofCompressorConfig, FriProverConfig,
         FriProverGatewayConfig, FriWitnessGeneratorConfig, FriWitnessVectorGeneratorConfig,
-        L1Secrets, ObservabilityConfig, PrometheusConfig, ProofDataHandlerConfig, Secrets,
+        L1Secrets, ObservabilityConfig, PrometheusConfig, ProofDataHandlerConfig,
+        ProtectiveReadsWriterConfig, Secrets,
     },
     ApiConfig, ContractVerifierConfig, DBConfig, EthConfig, EthWatchConfig, GasAdjusterConfig,
     GenesisConfig, ObjectStoreConfig, PostgresConfig, SnapshotsCreatorConfig,
@@ -306,5 +307,6 @@ fn load_env_config() -> anyhow::Result<TempConfigStore> {
         object_store_config: ObjectStoreConfig::from_env().ok(),
         observability: ObservabilityConfig::from_env().ok(),
         snapshot_creator: SnapshotsCreatorConfig::from_env().ok(),
+        protective_reads_writer_config: ProtectiveReadsWriterConfig::from_env().ok(),
     })
 }

core/bin/zksync_server/src/node_builder.rs

@@ -37,6 +37,7 @@ use zksync_node_framework::{
             StateKeeperLayer,
         },
         tee_verifier_input_producer::TeeVerifierInputProducerLayer,
+        vm_runner::protective_reads::ProtectiveReadsWriterLayer,
         web3_api::{
             caches::MempoolCacheLayer,
             server::{Web3ServerLayer, Web3ServerOptionalConfig},
@@ -399,6 +400,17 @@ impl MainNodeBuilder {
         Ok(self)
     }
 
+    fn add_vm_runner_protective_reads_layer(mut self) -> anyhow::Result<Self> {
+        let protective_reads_writer_config =
+            try_load_config!(self.configs.protective_reads_writer_config);
+        self.node.add_layer(ProtectiveReadsWriterLayer::new(
+            protective_reads_writer_config,
+            self.genesis_config.l2_chain_id,
+        ));
+
+        Ok(self)
+    }
+
     pub fn build(mut self, mut components: Vec<Component>) -> anyhow::Result<ZkStackService> {
         // Add "base" layers (resources and helper tasks).
         self = self
@@ -480,6 +492,9 @@ impl MainNodeBuilder {
                 Component::CommitmentGenerator => {
                     self = self.add_commitment_generator_layer()?;
                 }
+                Component::VmRunnerProtectiveReads => {
+                    self = self.add_vm_runner_protective_reads_layer()?;
+                }
             }
         }
         Ok(self.node.build()?)

core/lib/config/src/configs/general.rs

@@ -3,6 +3,7 @@ use crate::{
         chain::{CircuitBreakerConfig, MempoolConfig, OperationsManagerConfig, StateKeeperConfig},
         fri_prover_group::FriProverGroupConfig,
         house_keeper::HouseKeeperConfig,
+        vm_runner::ProtectiveReadsWriterConfig,
         FriProofCompressorConfig, FriProverConfig, FriProverGatewayConfig,
         FriWitnessGeneratorConfig, FriWitnessVectorGeneratorConfig, ObservabilityConfig,
         PrometheusConfig, ProofDataHandlerConfig,
@@ -32,4 +33,5 @@ pub struct GeneralConfig {
     pub eth: Option<EthConfig>,
     pub snapshot_creator: Option<SnapshotsCreatorConfig>,
     pub observability: Option<ObservabilityConfig>,
+    pub protective_reads_writer_config: Option<ProtectiveReadsWriterConfig>,
 }

core/lib/config/src/configs/mod.rs

@@ -20,6 +20,7 @@ pub use self::{
     secrets::{DatabaseSecrets, L1Secrets, Secrets},
     snapshots_creator::SnapshotsCreatorConfig,
     utils::PrometheusConfig,
+    vm_runner::ProtectiveReadsWriterConfig,
 };
 
 pub mod api;
@@ -46,6 +47,7 @@ pub mod proof_data_handler;
 pub mod secrets;
 pub mod snapshots_creator;
 pub mod utils;
+pub mod vm_runner;
 pub mod wallets;
 
 const BYTES_IN_MEGABYTE: usize = 1_024 * 1_024;

core/lib/config/src/configs/vm_runner.rs

@@ -0,0 +1,16 @@
+use serde::Deserialize;
+
+#[derive(Debug, Deserialize, Clone, PartialEq, Default)]
+pub struct ProtectiveReadsWriterConfig {
+    /// Path to the RocksDB data directory that serves state cache.
+    #[serde(default = "ProtectiveReadsWriterConfig::default_protective_reads_db_path")]
+    pub protective_reads_db_path: String,
+    /// How many max batches should be processed at the same time.
+    pub protective_reads_window_size: u32,
+}
+
+impl ProtectiveReadsWriterConfig {
+    fn default_protective_reads_db_path() -> String {
+        "./db/protective_reads_writer".to_owned()
+    }
+}

core/lib/dal/migrations/20240522215934_add_vm_runner_protective_reads_table.down.sql

@@ -0,0 +1 @@
+DROP TABLE IF EXISTS vm_runner_protective_reads;

core/lib/dal/migrations/20240522215934_add_vm_runner_protective_reads_table.up.sql

@@ -0,0 +1,7 @@
+CREATE TABLE IF NOT EXISTS vm_runner_protective_reads
+(
+    l1_batch_number       BIGINT    NOT NULL PRIMARY KEY,
+    created_at            TIMESTAMP NOT NULL,
+    updated_at            TIMESTAMP NOT NULL,
+    time_taken            TIME
+);

core/lib/dal/src/lib.rs

@@ -23,7 +23,7 @@ use crate::{
     sync_dal::SyncDal, system_dal::SystemDal,
     tee_verifier_input_producer_dal::TeeVerifierInputProducerDal, tokens_dal::TokensDal,
     tokens_web3_dal::TokensWeb3Dal, transactions_dal::TransactionsDal,
-    transactions_web3_dal::TransactionsWeb3Dal,
+    transactions_web3_dal::TransactionsWeb3Dal, vm_runner_dal::VmRunnerDal,
 };
 
 pub mod blocks_dal;
@@ -55,6 +55,7 @@ pub mod tokens_dal;
 pub mod tokens_web3_dal;
 pub mod transactions_dal;
 pub mod transactions_web3_dal;
+pub mod vm_runner_dal;
 
 #[cfg(test)]
 mod tests;
@@ -119,6 +120,8 @@ where
     fn snapshot_recovery_dal(&mut self) -> SnapshotRecoveryDal<'_, 'a>;
 
     fn pruning_dal(&mut self) -> PruningDal<'_, 'a>;
+
+    fn vm_runner_dal(&mut self) -> VmRunnerDal<'_, 'a>;
 }
 
 #[derive(Clone, Debug)]
@@ -229,4 +232,8 @@ impl<'a> CoreDal<'a> for Connection<'a, Core> {
     fn pruning_dal(&mut self) -> PruningDal<'_, 'a> {
         PruningDal { storage: self }
     }
+
+    fn vm_runner_dal(&mut self) -> VmRunnerDal<'_, 'a> {
+        VmRunnerDal { storage: self }
+    }
 }

core/lib/dal/src/vm_runner_dal.rs

@@ -0,0 +1,83 @@
+use zksync_db_connection::{connection::Connection, error::DalResult, instrument::InstrumentExt};
+use zksync_types::L1BatchNumber;
+
+use crate::Core;
+
+#[derive(Debug)]
+pub struct VmRunnerDal<'c, 'a> {
+    pub(crate) storage: &'c mut Connection<'a, Core>,
+}
+
+impl VmRunnerDal<'_, '_> {
+    pub async fn get_protective_reads_latest_processed_batch(
+        &mut self,
+    ) -> DalResult<L1BatchNumber> {
+        let row = sqlx::query!(
+            r#"
+            SELECT
+                COALESCE(MAX(l1_batch_number), 0) AS "last_processed_l1_batch!"
+            FROM
+                vm_runner_protective_reads
+            "#
+        )
+        .instrument("get_protective_reads_latest_processed_batch")
+        .report_latency()
+        .fetch_one(self.storage)
+        .await?;
+        Ok(L1BatchNumber(row.last_processed_l1_batch as u32))
+    }
+
+    pub async fn get_protective_reads_last_ready_batch(
+        &mut self,
+        window_size: u32,
+    ) -> DalResult<L1BatchNumber> {
+        let row = sqlx::query!(
+            r#"
+            WITH
+                available_batches AS (
+                    SELECT
+                        MAX(number) AS "last_batch"
+                    FROM
+                        l1_batches
+                ),
+                processed_batches AS (
+                    SELECT
+                        COALESCE(MAX(l1_batch_number), 0) + $1 AS "last_ready_batch"
+                    FROM
+                        vm_runner_protective_reads
+                )
+            SELECT
+                LEAST(last_batch, last_ready_batch) AS "last_ready_batch!"
+            FROM
+                available_batches
+                FULL JOIN processed_batches ON TRUE
+            "#,
+            window_size as i32
+        )
+        .instrument("get_protective_reads_last_ready_batch")
+        .report_latency()
+        .fetch_one(self.storage)
+        .await?;
+        Ok(L1BatchNumber(row.last_ready_batch as u32))
+    }
+
+    pub async fn mark_protective_reads_batch_as_completed(
+        &mut self,
+        l1_batch_number: L1BatchNumber,
+    ) -> DalResult<()> {
+        sqlx::query!(
+            r#"
+            INSERT INTO
+                vm_runner_protective_reads (l1_batch_number, created_at, updated_at)
+            VALUES
+                ($1, NOW(), NOW())
+            "#,
+            i64::from(l1_batch_number.0),
+        )
+        .instrument("mark_protective_reads_batch_as_completed")
+        .report_latency()
+        .execute(self.storage)
+        .await?;
+        Ok(())
+    }
+}

core/lib/env_config/src/lib.rs

@@ -24,6 +24,7 @@ mod utils;
 mod genesis;
 #[cfg(test)]
 mod test_utils;
+mod vm_runner;
 mod wallets;
 
 pub trait FromEnv: Sized {

core/lib/env_config/src/vm_runner.rs

@@ -0,0 +1,9 @@
+use zksync_config::configs::ProtectiveReadsWriterConfig;
+
+use crate::{envy_load, FromEnv};
+
+impl FromEnv for ProtectiveReadsWriterConfig {
+    fn from_env() -> anyhow::Result<Self> {
+        envy_load("vm_runner.protective_reads", "VM_RUNNER_PROTECTIVE_READS_")
+    }
+}

core/lib/protobuf_config/src/general.rs

@@ -37,6 +37,8 @@ impl ProtoRepr for proto::GeneralConfig {
             snapshot_creator: read_optional_repr(&self.snapshot_creator)
                 .context("snapshot_creator")?,
             observability: read_optional_repr(&self.observability).context("observability")?,
+            protective_reads_writer_config: read_optional_repr(&self.protective_reads_writer)
+                .context("vm_runner")?,
         })
     }
 
@@ -68,6 +70,10 @@ impl ProtoRepr for proto::GeneralConfig {
             eth: this.eth.as_ref().map(ProtoRepr::build),
             snapshot_creator: this.snapshot_creator.as_ref().map(ProtoRepr::build),
             observability: this.observability.as_ref().map(ProtoRepr::build),
+            protective_reads_writer: this
+                .protective_reads_writer_config
+                .as_ref()
+                .map(ProtoRepr::build),
         }
     }
 }

core/lib/protobuf_config/src/lib.rs

@@ -27,6 +27,7 @@ pub mod testonly;
 #[cfg(test)]
 mod tests;
 mod utils;
+mod vm_runner;
 mod wallets;
 
 use std::str::FromStr;

core/lib/protobuf_config/src/proto/config/general.proto

@@ -13,6 +13,7 @@ import "zksync/config/house_keeper.proto";
 import "zksync/config/observability.proto";
 import "zksync/config/snapshots_creator.proto";
 import "zksync/config/utils.proto";
+import "zksync/config/vm_runner.proto";
 
 message GeneralConfig {
   optional config.database.Postgres postgres = 1;
@@ -35,4 +36,5 @@ message GeneralConfig {
   optional config.prover.ProverGateway prover_gateway = 30;
   optional config.snapshot_creator.SnapshotsCreator snapshot_creator = 31;
   optional config.observability.Observability observability = 32;
+  optional config.vm_runner.ProtectiveReadsWriter protective_reads_writer = 33;
 }

core/lib/protobuf_config/src/proto/config/vm_runner.proto

@@ -0,0 +1,8 @@
+syntax = "proto3";
+
+package zksync.config.vm_runner;
+
+message ProtectiveReadsWriter {
+  optional string protective_reads_db_path = 1; // required; fs path
+  optional uint64 protective_reads_window_size = 2; // required
+}