fix(drive): uncommitted state if db transaction fails

packages/rs-drive-abci/src/abci/handler/finalize_block.rs

@@ -5,6 +5,7 @@ use crate::execution::types::block_execution_context::v0::BlockExecutionContextV
 use crate::platform_types::cleaned_abci_messages::finalized_block_cleaned_request::v0::FinalizeBlockCleanedRequest;
 use crate::platform_types::platform_state::v0::PlatformStateV0Methods;
 use crate::rpc::core::CoreRPCLike;
+use dpp::dashcore::Network;
 use std::sync::atomic::Ordering;
 use tenderdash_abci::proto::abci as proto;
 
@@ -66,7 +67,20 @@ where
         ));
     }
 
-    app.commit_transaction(platform_version)?;
+    // We had a chain halt on mainnet on block 32326. Compaction happened
+    // and transaction.commit() returned an error. Due to a bug in tenderdash,
+    // validators just proceeded on next block without committing data but keeping
+    // updated cache. To keep consistency with mainnet chain we have to skip
+    // commit of this block now on.
+    // TODO: verify that chain id is evo1
+    if !(app.platform().config.network == Network::Dash && block_height == 32326) {
+        // This is simplified solution until we have a better way to handle
+        // We still have caches in memory that corresponds to the data that
+        // we weren't able to commit. Solution is to restart the Drive, so all caches
+        // will be restored from the disk and try to process this block again
+        app.commit_transaction(platform_version)
+            .expect("commit transaction");
+    }
 
     app.platform()
         .committed_block_height_guard

packages/rs-drive-abci/src/abci/handler/info.rs

@@ -3,6 +3,7 @@ use crate::abci::AbciError;
 use crate::error::Error;
 use crate::platform_types::platform_state::v0::PlatformStateV0Methods;
 use crate::rpc::core::CoreRPCLike;
+use dpp::dashcore::Network;
 use dpp::version::DESIRED_PLATFORM_VERSION;
 use tenderdash_abci::proto::abci as proto;
 
@@ -21,28 +22,63 @@ where
 
     let platform_state = app.platform().state.load();
 
-    let state_app_hash = platform_state
+    let last_block_height = platform_state.last_committed_block_height() as i64;
+
+    // Verify that Platform State corresponds to Drive commited state
+    let drive_storage_root_hash = platform_state
         .last_committed_block_app_hash()
-        .map(|app_hash| app_hash.to_vec())
         .unwrap_or_default();
 
+    let platform_state_app_hash = app
+        .platform()
+        .drive
+        .grove
+        .root_hash(
+            None,
+            &platform_state
+                .current_platform_version()?
+                .drive
+                .grove_version,
+        )
+        .unwrap()?;
+
+    // We had a chain halt on mainnet on block 32326. Compaction happened
+    // and transaction.commit() returned an error. Due to a bug in tenderdash,
+    // validators just proceeded on next block without committing data but keeping
+    // updated cache. To keep consistency with mainnet chain we allow app hashes to be
+    // different for this block.
+    // TODO: verify that chain id is evo1
+    #[allow(clippy::collapsible_if)]
+    if !(app.platform().config.network == Network::Dash && last_block_height == 32326) {
+        // App hash in memory must be equal to app hash on disk
+        if drive_storage_root_hash != platform_state_app_hash {
+            // We panic because we can't recover from this situation.
+            // Better to restart the Drive, so we might self-heal the node
+            // reloading state form the disk
+            panic!(
+                "drive and platform state app hash mismatch: drive_storage_root_hash: {:?}, platform_state_app_hash: {:?}",
+                drive_storage_root_hash, platform_state_app_hash
+            );
+        }
+    }
+
     let desired_protocol_version = DESIRED_PLATFORM_VERSION.protocol_version;
 
     let response = proto::ResponseInfo {
         data: "".to_string(),
         app_version: desired_protocol_version as u64,
-        last_block_height: platform_state.last_committed_block_height() as i64,
+        last_block_height,
         version: env!("CARGO_PKG_VERSION").to_string(),
-        last_block_app_hash: state_app_hash.clone(),
+        last_block_app_hash: platform_state_app_hash.to_vec(),
     };
 
     tracing::debug!(
         desired_protocol_version,
         software_version = env!("CARGO_PKG_VERSION"),
         block_version = request.block_version,
         p2p_version = request.p2p_version,
-        app_hash = hex::encode(state_app_hash),
-        height = platform_state.last_committed_block_height(),
+        app_hash = hex::encode(platform_state_app_hash),
+        last_block_height,
         "Handshake with consensus engine",
     );
 

packages/rs-drive-abci/src/abci/handler/prepare_proposal.rs

@@ -11,6 +11,7 @@ use crate::platform_types::platform_state::v0::PlatformStateV0Methods;
 use crate::platform_types::state_transitions_processing_result::StateTransitionExecutionResult;
 use crate::rpc::core::CoreRPCLike;
 use dpp::dashcore::hashes::Hash;
+use dpp::dashcore::Network;
 use dpp::version::TryIntoPlatformVersioned;
 use drive::grovedb_storage::Error::RocksDBError;
 use tenderdash_abci::proto::abci as proto;
@@ -35,6 +36,44 @@ where
 
     let platform_state = app.platform().state.load();
 
+    // Verify that Platform State corresponds to Drive commited state
+    let drive_storage_root_hash = platform_state
+        .last_committed_block_app_hash()
+        .unwrap_or_default();
+
+    let platform_state_app_hash = app
+        .platform()
+        .drive
+        .grove
+        .root_hash(
+            None,
+            &platform_state
+                .current_platform_version()?
+                .drive
+                .grove_version,
+        )
+        .unwrap()?;
+
+    // We had a chain halt on mainnet on block 32326. Compaction happened
+    // and transaction.commit() returned an error. Due to a bug in tenderdash,
+    // validators just proceeded on next block without committing data but keeping
+    // updated cache. To keep consistency with mainnet chain we allow app hashes to be
+    // different for this block.
+    // TODO: verify that chain id is evo1
+    #[allow(clippy::collapsible_if)]
+    if !(app.platform().config.network == Network::Dash && request.height == 32327) {
+        // App hash in memory must be equal to app hash on disk
+        if drive_storage_root_hash != platform_state_app_hash {
+            // We panic because we can't recover from this situation.
+            // Better to restart the Drive, so we might self-heal the node
+            // reloading state form the disk
+            panic!(
+                "drive and platform state app hash mismatch: drive_storage_root_hash: {:?}, platform_state_app_hash: {:?}",
+                drive_storage_root_hash, platform_state_app_hash
+            );
+        }
+    }
+
     let last_committed_core_height = platform_state.last_committed_core_height();
 
     let starting_platform_version = platform_state.current_platform_version()?;

packages/rs-drive-abci/src/abci/handler/process_proposal.rs

@@ -12,6 +12,7 @@ use crate::platform_types::block_execution_outcome;
 use crate::platform_types::platform_state::v0::PlatformStateV0Methods;
 use crate::platform_types::state_transitions_processing_result::StateTransitionExecutionResult;
 use crate::rpc::core::CoreRPCLike;
+use dpp::dashcore::Network;
 use dpp::version::TryIntoPlatformVersioned;
 use drive::grovedb_storage::Error::RocksDBError;
 use tenderdash_abci::proto::abci as proto;
@@ -179,6 +180,44 @@ where
 
     let platform_state = app.platform().state.load();
 
+    // Verify that Platform State corresponds to Drive commited state
+    let drive_storage_root_hash = platform_state
+        .last_committed_block_app_hash()
+        .unwrap_or_default();
+
+    let platform_state_app_hash = app
+        .platform()
+        .drive
+        .grove
+        .root_hash(
+            None,
+            &platform_state
+                .current_platform_version()?
+                .drive
+                .grove_version,
+        )
+        .unwrap()?;
+
+    // We had a chain halt on mainnet on block 32326. Compaction happened
+    // and transaction.commit() returned an error. Due to a bug in tenderdash,
+    // validators just proceeded on next block without committing data but keeping
+    // updated cache. To keep consistency with mainnet chain we allow app hashes to be
+    // different for this block.
+    // TODO: verify that chain id is evo1
+    #[allow(clippy::collapsible_if)]
+    if !(app.platform().config.network == Network::Dash && request.height == 32327) {
+        // App hash in memory must be equal to app hash on disk
+        if drive_storage_root_hash != platform_state_app_hash {
+            // We panic because we can't recover from this situation.
+            // Better to restart the Drive, so we might self-heal the node
+            // reloading state form the disk
+            panic!(
+                "drive and platform state app hash mismatch: drive_storage_root_hash: {:?}, platform_state_app_hash: {:?}",
+                drive_storage_root_hash, platform_state_app_hash
+            );
+        }
+    }
+
     let starting_platform_version = platform_state.current_platform_version()?;
 
     // Running the proposal executes all the state transitions for the block

packages/rs-drive-abci/src/config.rs

@@ -233,8 +233,6 @@ struct PlatformConfigIntermediate {
     #[serde(flatten)]
     pub instant_lock: InstantLockConfig,
     pub block_spacing_ms: u64,
-    #[serde(default = "PlatformConfig::default_initial_protocol_version")]
-    pub initial_protocol_version: ProtocolVersion,
     pub db_path: PathBuf,
     #[serde(default)]
     pub rejections_path: Option<PathBuf>,
@@ -624,10 +622,6 @@ impl ExecutionConfig {
 }
 
 impl PlatformConfig {
-    fn default_initial_protocol_version() -> ProtocolVersion {
-        INITIAL_PROTOCOL_VERSION
-    }
-
     fn default_network() -> Network {
         Network::Dash
     }