diff --git a/Cargo.lock b/Cargo.lock index 6de9a8da..3af8db13 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -1349,6 +1349,25 @@ dependencies = [ "rand_core 0.10.1", ] +[[package]] +name = "chain_consistency" +version = "0.1.0" +dependencies = [ + "anyhow", + "borsh", + "common", + "futures", + "lee", + "lee_core", + "log", + "logos-blockchain-core", + "logos-blockchain-zone-sdk", + "serde", + "serde_json", + "thiserror 2.0.18", + "tokio", +] + [[package]] name = "chkstk_stub" version = "0.1.0" @@ -3850,6 +3869,7 @@ dependencies = [ "arc-swap", "async-stream", "borsh", + "chain_consistency", "common", "futures", "humantime-serde", @@ -3863,6 +3883,7 @@ dependencies = [ "storage", "tempfile", "testnet_initial_state", + "thiserror 2.0.18", "tokio", "url", ] @@ -3910,6 +3931,7 @@ dependencies = [ "base64", "common", "hex", + "indexer_core", "lee", "lee_core", "schemars 1.2.1", @@ -8890,6 +8912,7 @@ dependencies = [ "borsh", "bridge_core", "bytesize", + "chain_consistency", "chrono", "common", "faucet_core", diff --git a/Cargo.toml b/Cargo.toml index 88fe0c32..c7f07718 100644 --- a/Cargo.toml +++ b/Cargo.toml @@ -32,6 +32,7 @@ members = [ "lez/wallet", "lez/wallet-ffi", "lez/common", + "lez/chain_consistency", "lez/programs", "lez/programs/amm", "lez/programs/associated_token_account", @@ -64,6 +65,7 @@ members = [ lee = { path = "lee/state_machine" } lee_core = { path = "lee/state_machine/core" } common = { path = "lez/common" } +chain_consistency = { path = "lez/chain_consistency" } mempool = { path = "lez/mempool" } storage = { path = "lez/storage" } key_protocol = { path = "lee/key_protocol" } diff --git a/integration_tests/tests/indexer_stall.rs b/integration_tests/tests/indexer_stall.rs new file mode 100644 index 00000000..ae1b8b6a --- /dev/null +++ b/integration_tests/tests/indexer_stall.rs @@ -0,0 +1,54 @@ +#![expect( + clippy::tests_outside_test_module, + reason = "We don't care about these in tests" +)] + +use std::time::Duration; + +use anyhow::{Context as _, Result}; +use indexer_service_protocol::IndexerSyncState; +use indexer_service_rpc::RpcClient as _; +use integration_tests::{TestContext, wait_for_indexer_to_catch_up}; +use log::info; + +const CAUGHT_UP_STATUS_TIMEOUT: Duration = Duration::from_secs(60); + +/// Test that the indexer status RPC reports caught-up with no stall after a clean run. +/// +/// The sequencer keeps producing blocks while we assert, so the status is polled until a +/// `CaughtUp` snapshot is observed and the indexed tip is checked as a lower bound. +/// +/// TODO: Integration-level park testing (publishing a bad block to force a stall) is a follow-up +/// needing fault injection support in the test harness. +#[tokio::test] +async fn indexer_status_rpc_reports_caught_up_with_no_stall() -> Result<()> { + let ctx = TestContext::new().await?; + + let indexer_tip = wait_for_indexer_to_catch_up(&ctx).await?; + + let status = tokio::time::timeout(CAUGHT_UP_STATUS_TIMEOUT, async { + loop { + let status = ctx.indexer_client().get_status().await?; + if status.state == IndexerSyncState::CaughtUp { + return anyhow::Ok(status); + } + info!("Waiting for caught-up indexer status, got {status:?}"); + tokio::time::sleep(Duration::from_millis(500)).await; + } + }) + .await + .context("Timed out waiting for indexer status to report caught-up")??; + + assert!( + status.stall_reason.is_none(), + "indexer should have no stall reason after a clean run, got {status:?}" + ); + // test for >= here because the sequencer keeps producing blocks while we assert, + // so the indexed tip may be ahead of the tip we observed when we waited for caught-up. + assert!( + status.indexed_block_id >= Some(indexer_tip), + "status indexed_block_id should be at least the caught-up tip {indexer_tip}, got {status:?}" + ); + + Ok(()) +} diff --git a/lez/chain_consistency/Cargo.toml b/lez/chain_consistency/Cargo.toml new file mode 100644 index 00000000..02bf0ea0 --- /dev/null +++ b/lez/chain_consistency/Cargo.toml @@ -0,0 +1,26 @@ +[package] +name = "chain_consistency" +version = "0.1.0" +edition = "2024" +license.workspace = true + +[lints] +workspace = true + +[dependencies] +common.workspace = true +lee.workspace = true +lee_core.workspace = true + +logos-blockchain-zone-sdk.workspace = true +logos-blockchain-core.workspace = true +anyhow.workspace = true +borsh.workspace = true +futures.workspace = true +log.workspace = true +serde.workspace = true +thiserror.workspace = true +tokio.workspace = true + +[dev-dependencies] +serde_json.workspace = true diff --git a/lez/chain_consistency/src/apply.rs b/lez/chain_consistency/src/apply.rs new file mode 100644 index 00000000..e90d21a3 --- /dev/null +++ b/lez/chain_consistency/src/apply.rs @@ -0,0 +1,182 @@ +//! Validating and applying channel L2 blocks to a `V03State`. +//! +//! These primitives are shared by the consumers that reconstruct L2 state from +//! the Bedrock channel. + +use common::{ + HashType, + block::Block, + transaction::{LeeTransaction, clock_invocation}, +}; +use lee::{GENESIS_BLOCK_ID, V03State}; +use lee_core::BlockId; +use serde::{Deserialize, Serialize}; + +/// Why a channel L2 block could not be validated or applied. +/// +/// Persisted in `RocksDB` (via the Indexer's stall reason), so every variant +/// must be `Clone + Serialize + Deserialize`. +#[derive(Debug, Clone, Serialize, Deserialize, thiserror::Error)] +pub enum BlockIngestError { + #[error("Failed to deserialize L2 block: {0}")] + /// Here we store the error string that is derived from [`borsh::from_slice`]'s [`Err`]. + Deserialize(String), + #[error("Unexpected block id: expected {expected}, got {got}")] + UnexpectedBlockId { expected: BlockId, got: BlockId }, + #[error("Broken chain link: expected prev {expected_prev}, got {got_prev}")] + BrokenChainLink { + expected_prev: HashType, + got_prev: HashType, + }, + #[error("Block hash mismatch: computed {computed}, header {header}")] + HashMismatch { + computed: HashType, + header: HashType, + }, + #[error("Block has no transactions")] + EmptyBlock, + #[error("Last transaction must be the public clock invocation for the block timestamp")] + InvalidClockTransaction, + #[error("Genesis block must contain only public transactions")] + NonPublicGenesisTransaction, + #[error("State transition failed at transaction {tx_index}: {reason}")] + StateTransition { + /// Index of the failing transaction within the block body. + tx_index: u64, + /// Reason string from `lee::Error` to `anyhow::Error` to `{:#}`. + /// + /// This is required because `lee::Error` is not `Clone + Serialize + Deserialize`, so we + /// cannot store it directly. + reason: String, + }, +} + +/// The last successfully applied block a candidate must extend. +pub struct Tip { + pub block_id: BlockId, + pub hash: HashType, +} + +/// Checks that `block` is the valid continuation of `tip`: hash integrity, +/// then block-id continuity, then `prev_block_hash` linkage. A `None` tip +/// (cold store) expects the genesis block. +pub fn validate_against_tip(tip: Option<&Tip>, block: &Block) -> Result<(), BlockIngestError> { + let computed = block.recompute_hash(); + if computed != block.header.hash { + return Err(BlockIngestError::HashMismatch { + computed, + header: block.header.hash, + }); + } + + match tip { + None => { + if block.header.block_id != GENESIS_BLOCK_ID { + return Err(BlockIngestError::UnexpectedBlockId { + expected: GENESIS_BLOCK_ID, + got: block.header.block_id, + }); + } + } + Some(tip) => { + let expected = tip + .block_id + .checked_add(1) + .expect("block id should not overflow"); + if block.header.block_id != expected { + return Err(BlockIngestError::UnexpectedBlockId { + expected, + got: block.header.block_id, + }); + } + if block.header.prev_block_hash != tip.hash { + return Err(BlockIngestError::BrokenChainLink { + expected_prev: tip.hash, + got_prev: block.header.prev_block_hash, + }); + } + } + } + Ok(()) +} + +/// Applies a block's transactions to `state`. +pub fn apply_block(block: &Block, state: &mut V03State) -> Result<(), BlockIngestError> { + let (clock_tx, user_txs) = block + .body + .transactions + .split_last() + .ok_or(BlockIngestError::EmptyBlock)?; + + let expected_clock = LeeTransaction::Public(clock_invocation(block.header.timestamp)); + if *clock_tx != expected_clock { + return Err(BlockIngestError::InvalidClockTransaction); + } + + let is_genesis = block.header.block_id == GENESIS_BLOCK_ID; + for (tx_index, transaction) in user_txs.iter().enumerate() { + let state_transition = |err: anyhow::Error| BlockIngestError::StateTransition { + tx_index: tx_index.try_into().expect("tx index fits in u64"), + reason: format!("{err:#}"), + }; + if is_genesis { + let LeeTransaction::Public(public_tx) = transaction else { + return Err(BlockIngestError::NonPublicGenesisTransaction); + }; + state + .transition_from_public_transaction( + public_tx, + block.header.block_id, + block.header.timestamp, + ) + .map_err(|err| state_transition(err.into()))?; + } else { + transaction + .clone() + .execute_on_state(state, block.header.block_id, block.header.timestamp) + .map_err(|err| state_transition(err.into()))?; + } + } + + let LeeTransaction::Public(clock_public_tx) = clock_tx else { + return Err(BlockIngestError::InvalidClockTransaction); + }; + state + .transition_from_public_transaction( + clock_public_tx, + block.header.block_id, + block.header.timestamp, + ) + .map_err(|err| BlockIngestError::StateTransition { + tx_index: user_txs.len().try_into().expect("tx index fits in u64"), + reason: format!("{:#}", anyhow::Error::from(err)), + })?; + + Ok(()) +} + +#[cfg(test)] +mod tests { + use super::*; + + #[test] + fn serializes_and_round_trips_externally_tagged() { + let err = BlockIngestError::UnexpectedBlockId { + expected: 5, + got: 7, + }; + let value = serde_json::to_value(&err).expect("serialize"); + assert_eq!( + value, + serde_json::json!({ "UnexpectedBlockId": { "expected": 5, "got": 7 } }) + ); + let back: BlockIngestError = serde_json::from_value(value).expect("deserialize"); + assert!(matches!( + back, + BlockIngestError::UnexpectedBlockId { + expected: 5, + got: 7 + } + )); + } +} diff --git a/lez/chain_consistency/src/consistency.rs b/lez/chain_consistency/src/consistency.rs new file mode 100644 index 00000000..dce720b1 --- /dev/null +++ b/lez/chain_consistency/src/consistency.rs @@ -0,0 +1,429 @@ +//! Startup check that a local store still belongs to the chain the connected +//! channel serves. + +use anyhow::Result; +use common::{HashType, block::Block}; +use futures::StreamExt as _; +use lee_core::BlockId; +use log::warn; +use logos_blockchain_core::mantle::ops::channel::ChannelId; +use logos_blockchain_zone_sdk::{Slot, ZoneMessage, adapter, indexer::ZoneIndexer}; + +/// Upper bound on the channel reads of the startup consistency check. +const CHANNEL_READ_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(60); + +/// Result of comparing a caller's stored chain against the channel. +pub enum ChainConsistency { + /// Channel still serves our anchor block (the stored tip position, or the + /// parked block while stalled). + Consistent, + /// We could not determine the outcome due to one of: + /// + /// - cold store (no anchor to compare at) + /// - the channel served only blocks newer than the anchor + /// - or the channel read was inconclusive (timeout / error / empty stream) + /// + /// NOTE: None of these prove a reset, so the caller proceeds. + /// A genuine divergence is still caught later when applying the channel history. + Inconclusive, + /// Positive evidence that the channel is a different chain than the store. + /// + /// Details in [`ChainMismatch`], and impl's Display trait. + Inconsistent(ChainMismatch), +} + +/// The evidence behind a [`ChainConsistency::Inconsistent`]. +pub enum ChainMismatch { + /// The channel serves a different block at the anchor's id. + Block { + ours: (BlockId, HashType), + channel: (BlockId, HashType), + }, + /// The channel serves a block at/below the anchor's id past the anchor + /// slot; on the same chain those ids live at earlier slots. + ReinscribedBlock { + channel: (BlockId, HashType), + slot: Slot, + anchor_slot: Slot, + }, + /// The channel has content past the anchor slot but no longer the + /// inscription we anchored on. + AnchorSlotChanged { anchor_slot: Slot }, + /// The channel does not exist on the connected chain. + ChannelMissing, + /// The channel's history ends before the anchor slot. + ChannelBehindAnchor { tip_slot: Slot, anchor_slot: Slot }, +} + +impl std::fmt::Display for ChainMismatch { + fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + match self { + Self::Block { ours, channel } => write!( + f, + "stored block {} {} != channel block {} {}", + ours.0, ours.1, channel.0, channel.1 + ), + Self::ReinscribedBlock { + channel, + slot, + anchor_slot, + } => write!( + f, + "channel re-serves block {} {} at slot {} past our anchor slot {}", + channel.0, + channel.1, + slot.into_inner(), + anchor_slot.into_inner() + ), + Self::AnchorSlotChanged { anchor_slot } => write!( + f, + "channel content at slot {} no longer includes the inscription we parked on", + anchor_slot.into_inner() + ), + Self::ChannelMissing => write!(f, "channel does not exist on the connected chain"), + Self::ChannelBehindAnchor { + tip_slot, + anchor_slot, + } => write!( + f, + "channel tip slot {} is behind our anchor slot {}", + tip_slot.into_inner(), + anchor_slot.into_inner() + ), + } + } +} + +/// A block that must still be inscribed at `slot` if the channel is the chain +/// the store was built from: the tip at the read cursor, or the recorded +/// parked block while stalled. +pub struct Anchor { + slot: Slot, + /// The anchor block's `(id, hash)`. + /// + /// `None` when anchored on an undeserializable inscription (no header was recorded). + block: Option<(BlockId, HashType)>, +} + +impl Anchor { + /// Builds an anchor at `slot` on the block `(id, hash)`, or a headerless + /// anchor (`None`) when only the slot is known. + #[must_use] + pub const fn new(slot: Slot, block: Option<(BlockId, HashType)>) -> Self { + Self { slot, block } + } + + /// Probes a channel message read at/after the anchor slot. + /// See [`verify_chain_consistency`]. + fn probe_anchor_slot(&self, msg: &ZoneMessage, slot: Slot) -> AnchorProbe { + if slot < self.slot { + return AnchorProbe::KeepLooking; + } + let Some((anchor_id, anchor_hash)) = self.block else { + // Anchored on an undeserializable inscription: any message still + // present at that slot means the history is intact. + return if slot == self.slot { + AnchorProbe::SameChain + } else { + AnchorProbe::Mismatch(ChainMismatch::AnchorSlotChanged { + anchor_slot: self.slot, + }) + }; + }; + let ZoneMessage::Block(zone_block) = msg else { + return AnchorProbe::KeepLooking; + }; + let Ok(block) = borsh::from_slice::(&zone_block.data) else { + return AnchorProbe::KeepLooking; + }; + let (id, hash) = (block.header.block_id, block.header.hash); + if id == anchor_id { + return if hash == anchor_hash { + AnchorProbe::SameChain + } else { + AnchorProbe::Mismatch(ChainMismatch::Block { + ours: (anchor_id, anchor_hash), + channel: (id, hash), + }) + }; + } + if id > anchor_id { + return AnchorProbe::Bail; + } + if slot == self.slot { + // Older ids can share the anchor's slot on the same chain. + return AnchorProbe::KeepLooking; + } + // An id below the anchor served past the anchor slot is impossible on the + // same chain, even if the content is identical (deterministic genesis). + AnchorProbe::Mismatch(ChainMismatch::ReinscribedBlock { + channel: (id, hash), + slot, + anchor_slot: self.slot, + }) + } +} + +/// What a single channel message tells the anchored consistency check. +enum AnchorProbe { + /// The anchor is still in place: same chain. + SameChain, + Mismatch(ChainMismatch), + /// Only newer ids past the anchor: plausible on the same chain, so stop + /// scanning without a verdict. + Bail, + KeepLooking, +} + +/// Detects when a local store belongs to a different chain than the connected +/// L1 (e.g. a wiped/restarted Bedrock) so startup can react instead of silently +/// diverging. +/// +/// Verifies the channel still carries the anchor block at its slot. The anchor +/// was finalized at `anchor.slot`, so the same chain must still serve it there, +/// while a reset chain re-inscribes its content only at later wall-clock slots. +/// Only positive evidence of a different chain yields +/// [`ChainConsistency::Inconsistent`]; absence of data stays +/// [`ChainConsistency::Inconclusive`]. +/// +/// `node` need only implement the zone-sdk [`adapter::Node`] trait; a throwaway +/// [`ZoneIndexer`] is built internally for the channel read. +pub async fn verify_chain_consistency( + node: &N, + channel_id: ChannelId, + anchor: &Anchor, +) -> Result +where + N: adapter::Node + Clone + Sync, +{ + match node.channel_state(channel_id).await { + Ok(state) => { + if let Some(mismatch) = frontier_verdict(anchor.slot, state.map(|s| s.tip_slot)) { + return Ok(ChainConsistency::Inconsistent(mismatch)); + } + } + Err(err) => { + warn!("Failed to read channel state for the consistency check: {err:#}"); + } + } + + // `next_messages` is exclusive, so `slot - 1` includes the anchor slot. + let Some(from_slot) = anchor.slot.into_inner().checked_sub(1) else { + return Ok(ChainConsistency::Inconclusive); + }; + + let zone_indexer = ZoneIndexer::new(channel_id, node.clone()); + let scan = async { + let stream = zone_indexer + .next_messages(Some(Slot::from(from_slot))) + .await?; + let mut stream = std::pin::pin!(stream); + + while let Some((msg, slot)) = stream.next().await { + match anchor.probe_anchor_slot(&msg, slot) { + AnchorProbe::SameChain => return Ok(Some(ChainConsistency::Consistent)), + AnchorProbe::Mismatch(mismatch) => { + return Ok(Some(ChainConsistency::Inconsistent(mismatch))); + } + AnchorProbe::Bail => return Ok(Some(ChainConsistency::Inconclusive)), + AnchorProbe::KeepLooking => { /* dont do anything */ } + } + } + Ok::<_, anyhow::Error>(None) + }; + + match tokio::time::timeout(CHANNEL_READ_TIMEOUT, scan).await { + Ok(Ok(Some(outcome))) => Ok(outcome), + Ok(Ok(None)) => Ok(ChainConsistency::Inconclusive), + Ok(Err(err)) => { + warn!("Failed to read the anchor slot for the consistency check; proceeding: {err:#}"); + Ok(ChainConsistency::Inconclusive) + } + Err(_elapsed) => { + warn!("Timed out reading the anchor slot for the consistency check; proceeding"); + Ok(ChainConsistency::Inconclusive) + } + } +} + +/// Classifies an already-fetched slice of channel messages against `anchor`. +/// +/// The message-slice counterpart of [`verify_chain_consistency`], for callers +/// that read the channel history themselves (e.g. a sequencer that reads once +/// and both verifies and reconstructs from the same messages). `messages` must +/// be the finalized channel history from the anchor slot onward, in slot order. +/// +/// Reuses the same [`Anchor::probe_anchor_slot`] and [`frontier_verdict`] logic +/// as the streaming path, so both consumers share one definition of a mismatch. +#[must_use] +pub fn classify_channel( + anchor: &Anchor, + channel_tip_slot: Option, + messages: &[(ZoneMessage, Slot)], +) -> ChainConsistency { + if let Some(mismatch) = frontier_verdict(anchor.slot, channel_tip_slot) { + return ChainConsistency::Inconsistent(mismatch); + } + for (msg, slot) in messages { + match anchor.probe_anchor_slot(msg, *slot) { + AnchorProbe::SameChain => return ChainConsistency::Consistent, + AnchorProbe::Mismatch(mismatch) => return ChainConsistency::Inconsistent(mismatch), + AnchorProbe::Bail => return ChainConsistency::Inconclusive, + AnchorProbe::KeepLooking => {} + } + } + ChainConsistency::Inconclusive +} + +/// Checks the channel frontier against the anchor slot. +/// +/// The anchor block was finalized at `anchor_slot`, so on the same chain the +/// channel tip can never be behind it, and the channel must exist. +fn frontier_verdict(anchor_slot: Slot, channel_tip_slot: Option) -> Option { + match channel_tip_slot { + None => Some(ChainMismatch::ChannelMissing), + Some(tip_slot) if tip_slot < anchor_slot => Some(ChainMismatch::ChannelBehindAnchor { + tip_slot, + anchor_slot, + }), + Some(_) => None, + } +} + +#[cfg(test)] +mod tests { + use common::block::HashableBlockData; + use logos_blockchain_core::mantle::ops::channel::{MsgId, inscribe::Inscription}; + use logos_blockchain_zone_sdk::ZoneBlock; + + use super::*; + + fn test_block(block_id: BlockId, timestamp: u64) -> Block { + HashableBlockData { + block_id, + prev_block_hash: HashType([0; 32]), + timestamp, + transactions: vec![], + } + .into_pending_block(&lee::PrivateKey::try_new([7; 32]).expect("valid key")) + } + + fn block_msg(block: &Block) -> ZoneMessage { + let bytes = borsh::to_vec(block).expect("serialize"); + ZoneMessage::Block(ZoneBlock { + id: MsgId::from([0_u8; 32]), + data: Inscription::try_from(bytes.as_slice()).expect("inscription"), + }) + } + + fn anchor_for(block: &Block, slot: Slot) -> Anchor { + Anchor::new(slot, Some((block.header.block_id, block.header.hash))) + } + + #[test] + fn probe_finds_anchor_block_at_slot() { + let tip = test_block(5, 42); + let anchor = anchor_for(&tip, Slot::from(1_000)); + assert!(matches!( + anchor.probe_anchor_slot(&block_msg(&tip), Slot::from(1_000)), + AnchorProbe::SameChain + )); + } + + #[test] + fn probe_flags_different_block_at_anchor_id() { + let tip = test_block(5, 42); + let anchor = anchor_for(&tip, Slot::from(1_000)); + // Same id, different content (timestamp) => different hash. + let other = test_block(5, 43); + assert!(matches!( + anchor.probe_anchor_slot(&block_msg(&other), Slot::from(1_000)), + AnchorProbe::Mismatch(ChainMismatch::Block { .. }) + )); + } + + #[test] + fn probe_flags_old_id_reinscribed_past_the_anchor_slot() { + // A reset chain re-inscribes from genesis at later slots. Even if the + // content is byte-identical (deterministic genesis), an id at/below + // the anchor past the anchor slot is impossible on the same chain. + let tip = test_block(5, 42); + let anchor = anchor_for(&tip, Slot::from(1_000)); + let genesis = test_block(1, 0); + assert!(matches!( + anchor.probe_anchor_slot(&block_msg(&genesis), Slot::from(1_001)), + AnchorProbe::Mismatch(ChainMismatch::ReinscribedBlock { .. }) + )); + } + + #[test] + fn probe_skips_older_blocks_sharing_the_anchor_slot() { + let tip = test_block(5, 42); + let anchor = anchor_for(&tip, Slot::from(1_000)); + let earlier = test_block(4, 41); + assert!(matches!( + anchor.probe_anchor_slot(&block_msg(&earlier), Slot::from(1_000)), + AnchorProbe::KeepLooking + )); + } + + #[test] + fn probe_bails_on_newer_ids_past_the_anchor() { + // Blocks newer than the anchor are plausible on the same chain (e.g. + // published while we were down), so they must never count as evidence. + let tip = test_block(5, 42); + let anchor = anchor_for(&tip, Slot::from(1_000)); + let newer = test_block(6, 43); + assert!(matches!( + anchor.probe_anchor_slot(&block_msg(&newer), Slot::from(1_001)), + AnchorProbe::Bail + )); + } + + #[test] + fn probe_skips_undeserializable_inscriptions() { + let tip = test_block(5, 42); + let anchor = anchor_for(&tip, Slot::from(1_000)); + let garbage = ZoneMessage::Block(ZoneBlock { + id: MsgId::from([0_u8; 32]), + data: Inscription::try_from(&[1_u8, 2, 3][..]).expect("inscription"), + }); + assert!(matches!( + anchor.probe_anchor_slot(&garbage, Slot::from(1_000)), + AnchorProbe::KeepLooking + )); + } + + #[test] + fn probe_accepts_any_message_for_a_headerless_anchor() { + // A deserialize park records no header: any message still present at + // the anchor slot means the history is intact. + let anchor = Anchor::new(Slot::from(1_000), None); + let garbage = ZoneMessage::Block(ZoneBlock { + id: MsgId::from([0_u8; 32]), + data: Inscription::try_from(&[1_u8, 2, 3][..]).expect("inscription"), + }); + assert!(matches!( + anchor.probe_anchor_slot(&garbage, Slot::from(1_000)), + AnchorProbe::SameChain + )); + assert!(matches!( + anchor.probe_anchor_slot(&garbage, Slot::from(1_001)), + AnchorProbe::Mismatch(ChainMismatch::AnchorSlotChanged { .. }) + )); + } + + #[test] + fn frontier_flags_missing_channel_and_short_history() { + assert!(matches!( + frontier_verdict(Slot::from(1_000), None), + Some(ChainMismatch::ChannelMissing) + )); + assert!(matches!( + frontier_verdict(Slot::from(1_000), Some(Slot::from(999))), + Some(ChainMismatch::ChannelBehindAnchor { .. }) + )); + assert!(frontier_verdict(Slot::from(1_000), Some(Slot::from(1_000))).is_none()); + assert!(frontier_verdict(Slot::from(1_000), Some(Slot::from(2_000))).is_none()); + } +} diff --git a/lez/chain_consistency/src/lib.rs b/lez/chain_consistency/src/lib.rs new file mode 100644 index 00000000..7ef25725 --- /dev/null +++ b/lez/chain_consistency/src/lib.rs @@ -0,0 +1,9 @@ +//! Reconstructing and verifying L2 chain state from a Bedrock (L1) channel. + +pub use apply::{BlockIngestError, Tip, apply_block, validate_against_tip}; +pub use consistency::{ + Anchor, ChainConsistency, ChainMismatch, classify_channel, verify_chain_consistency, +}; + +pub mod apply; +pub mod consistency; diff --git a/lez/common/src/block.rs b/lez/common/src/block.rs index 6e956f9f..a8149e7c 100644 --- a/lez/common/src/block.rs +++ b/lez/common/src/block.rs @@ -55,6 +55,21 @@ pub struct Block { pub bedrock_status: BedrockStatus, } +impl Block { + /// Recomputes the hash from this block's contents, for integrity verification + /// against the value stored in `header.hash`. + #[must_use] + pub fn recompute_hash(&self) -> BlockHash { + HashableBlockData { + block_id: self.header.block_id, + prev_block_hash: self.header.prev_block_hash, + timestamp: self.header.timestamp, + transactions: self.body.transactions.clone(), + } + .compute_hash() + } +} + impl Serialize for Block { fn serialize(&self, serializer: S) -> Result { crate::borsh_base64::serialize(self, serializer) @@ -76,11 +91,13 @@ pub struct HashableBlockData { } impl HashableBlockData { + /// Domain-separated hash of the block contents: `SHA256(PREFIX || borsh(self))`. + /// The single source of truth for both producing and verifying a block hash. #[must_use] - pub fn into_pending_block(self, signing_key: &lee::PrivateKey) -> Block { + pub fn compute_hash(&self) -> BlockHash { const PREFIX: &[u8; 32] = b"/LEE/v0.3/Message/Block/\x00\x00\x00\x00\x00\x00\x00\x00"; - let data_bytes = borsh::to_vec(&self).unwrap(); + let data_bytes = borsh::to_vec(self).unwrap(); let mut bytes = Vec::with_capacity( PREFIX .len() @@ -89,8 +106,12 @@ impl HashableBlockData { ); bytes.extend_from_slice(PREFIX); bytes.extend_from_slice(&data_bytes); + OwnHasher::hash(&bytes) + } - let hash = OwnHasher::hash(&bytes); + #[must_use] + pub fn into_pending_block(self, signing_key: &lee::PrivateKey) -> Block { + let hash = self.compute_hash(); let signature = lee::Signature::new(signing_key, &hash.0); Block { header: BlockHeader { @@ -132,4 +153,33 @@ mod tests { let block_from_bytes = borsh::from_slice::(&bytes).unwrap(); assert_eq!(hashable, block_from_bytes); } + + #[test] + fn recompute_hash_matches_header_for_well_formed_block() { + let key = lee::PrivateKey::try_new([7_u8; 32]).expect("valid key"); + let block = HashableBlockData { + block_id: 5, + prev_block_hash: HashType([9_u8; 32]), + timestamp: 42, + transactions: vec![test_utils::produce_dummy_empty_transaction()], + } + .into_pending_block(&key); + assert_eq!(block.recompute_hash(), block.header.hash); + } + + #[test] + fn recompute_hash_detects_tampering() { + let key = lee::PrivateKey::try_new([7_u8; 32]).expect("valid key"); + let block = HashableBlockData { + block_id: 5, + prev_block_hash: HashType([9_u8; 32]), + timestamp: 42, + transactions: vec![test_utils::produce_dummy_empty_transaction()], + } + .into_pending_block(&key); + + let mut tampered = block; + tampered.header.timestamp = 99; // header changed; stale hash no longer matches + assert_ne!(tampered.recompute_hash(), tampered.header.hash); + } } diff --git a/lez/common/src/transaction.rs b/lez/common/src/transaction.rs index b5aee648..b5f0bc87 100644 --- a/lez/common/src/transaction.rs +++ b/lez/common/src/transaction.rs @@ -92,9 +92,8 @@ impl LeeTransaction { Ok(diff) } - /// Computes the validated state diff without enforcing the system-account - /// restriction. Shared by [`Self::validate_on_state`] and - /// [`Self::execute_without_system_accounts_check_on_state`]. + /// Computes the validated state diff. Shared by [`Self::validate_on_state`] + /// (which adds the system-account guards) and [`Self::execute_on_state`]. fn compute_state_diff( &self, state: &V03State, @@ -129,16 +128,12 @@ impl LeeTransaction { Ok(self) } - /// Similar to [`Self::execute_check_on_state`], but skips the system-account guard. + /// Executes the transaction against the current state and applies the resulting diff, + /// without the system-account guards enforced by [`Self::execute_check_on_state`]. /// - /// FIXME: HOT FIX (testnet v0.2): the indexer replays blocks the sequencer already - /// accepted, including sequencer-generated deposit transactions that - /// legitimately modify the bridge account. The `TransactionOrigin::Sequencer` - /// tag that lets the sequencer bypass the guard is not carried in the block, - /// so the indexer cannot yet distinguish deposit txs from user txs. - /// - /// REMOVE ME when the indexer can authenticate deposit transactions. - pub fn execute_without_system_accounts_check_on_state( + /// The indexer replays blocks the sequencer already validated and inscribed on Bedrock, + /// so it trusts those inscriptions and re-derives state without re-validating them. + pub fn execute_on_state( self, state: &mut V03State, block_id: BlockId, diff --git a/lez/configs/docker-all-in-one/indexer_config.json b/lez/configs/docker-all-in-one/indexer_config.json index c1ff65b0..5791f64a 100644 --- a/lez/configs/docker-all-in-one/indexer_config.json +++ b/lez/configs/docker-all-in-one/indexer_config.json @@ -3,5 +3,6 @@ "bedrock_config": { "addr": "http://logos-blockchain-node-0:18080" }, - "channel_id": "0101010101010101010101010101010101010101010101010101010101010101" + "channel_id": "0101010101010101010101010101010101010101010101010101010101010101", + "allow_chain_reset": true } diff --git a/lez/indexer/core/Cargo.toml b/lez/indexer/core/Cargo.toml index b49f6fee..7e1b2188 100644 --- a/lez/indexer/core/Cargo.toml +++ b/lez/indexer/core/Cargo.toml @@ -13,6 +13,7 @@ testnet = [] [dependencies] common.workspace = true +chain_consistency.workspace = true logos-blockchain-zone-sdk.workspace = true lee.workspace = true lee_core.workspace = true @@ -29,6 +30,7 @@ futures.workspace = true url.workspace = true logos-blockchain-core.workspace = true serde_json.workspace = true +thiserror.workspace = true async-stream.workspace = true tokio.workspace = true diff --git a/lez/indexer/core/src/block_store.rs b/lez/indexer/core/src/block_store.rs index 4d583686..17407745 100644 --- a/lez/indexer/core/src/block_store.rs +++ b/lez/indexer/core/src/block_store.rs @@ -1,18 +1,31 @@ use std::{path::Path, sync::Arc}; use anyhow::{Context as _, Result}; +use chain_consistency::{BlockIngestError, Tip}; use common::{ - block::{BedrockStatus, Block}, - transaction::{LeeTransaction, clock_invocation}, + block::{BedrockStatus, Block, BlockHeader}, + transaction::LeeTransaction, }; use lee::{Account, AccountId, V03State}; use lee_core::BlockId; -use log::info; +use log::warn; use logos_blockchain_core::header::HeaderId; use logos_blockchain_zone_sdk::Slot; use storage::indexer::RocksDBIO; use tokio::sync::RwLock; +use crate::stall_reason::StallReason; + +/// Outcome of feeding a parsed L2 block to the validated tip. +pub enum AcceptOutcome { + /// Chained and applied; tip and L1 read cursor both advance. + Applied, + /// A duplicate re-delivery of the current tip. Just L2 advances. + AlreadyApplied, + /// Did not chain or failed to apply; tip stays frozen, stall recorded. + Parked(BlockIngestError), +} + #[derive(Clone)] pub struct IndexerStore { dbio: Arc, @@ -118,6 +131,29 @@ impl IndexerStore { Ok(()) } + pub fn get_stall_reason(&self) -> Result> { + let Some(bytes) = self.dbio.get_stall_reason_bytes()? else { + return Ok(None); + }; + let stall: Option = + serde_json::from_slice(&bytes).context("Failed to deserialize stored stall reason")?; + Ok(stall) + } + + pub fn set_stall_reason(&self, stall: &Option) -> Result<()> { + let bytes = serde_json::to_vec(stall).context("Failed to serialize stall reason")?; + self.dbio.put_stall_reason_bytes(&bytes)?; + Ok(()) + } + + /// Clears a recorded stall marker if one is present, skipping the write otherwise. + fn clear_stall_if_present(&self) -> Result<()> { + if self.get_stall_reason()?.is_some() { + self.set_stall_reason(&None)?; + } + Ok(()) + } + /// Recalculation of final state directly from DB. /// /// Used for indexer healthcheck. @@ -139,137 +175,147 @@ impl IndexerStore { .get_account_by_id(*account_id)) } - pub async fn put_block(&self, mut block: Block, l1_header: HeaderId) -> Result<()> { - info!("Applying block {}", block.header.block_id); - { - let mut state_guard = self.current_state.write().await; + /// The last successfully applied block, or `None` on a cold store. + /// Read fresh from the store each call. + fn validated_tip(&self) -> Result> { + let Some(block_id) = self.dbio.get_meta_last_block_id_in_db()? else { + return Ok(None); + }; + let Some(block) = self.dbio.get_block(block_id)? else { + return Ok(None); + }; + Ok(Some(Tip { + block_id, + hash: block.header.hash, + })) + } - let (clock_tx, user_txs) = block - .body - .transactions - .split_last() - .ok_or_else(|| anyhow::anyhow!("Block has no transactions"))?; - - anyhow::ensure!( - *clock_tx == LeeTransaction::Public(clock_invocation(block.header.timestamp)), - "Last transaction in block must be the clock invocation for the block timestamp" - ); - - let is_genesis = block.header.block_id == 1; - for transaction in user_txs { - if is_genesis { - let genesis_tx = match transaction { - LeeTransaction::Public(public_tx) => public_tx, - LeeTransaction::PrivacyPreserving(_) - | LeeTransaction::ProgramDeployment(_) => { - anyhow::bail!("Genesis block should contain only public transactions") - } - }; - state_guard - .transition_from_public_transaction( - genesis_tx, - block.header.block_id, - block.header.timestamp, - ) - .context("Failed to execute genesis public transaction")?; - } else { - transaction - .clone() - .transaction_stateless_check()? - // FIXME: HOT FIX (testnet v0.2): does not check for system account updates due to - // sequencer-generated deposit tx'es; - // CHANGE ME back to `execute_check_on_state` when the indexer can authenticate deposit transactions - .execute_without_system_accounts_check_on_state( - &mut state_guard, - block.header.block_id, - block.header.timestamp, - )?; - } + /// Record the stall reason. + /// + /// - First stall is stored verbatim + /// - Subsequent stalls only bump `orphans_since`, preserving the original cause. + pub fn record_stall( + &self, + header: Option<&BlockHeader>, + l1_slot: Slot, + error: BlockIngestError, + ) -> Result<()> { + let stall = match self.get_stall_reason()? { + Some(mut existing) => { + existing.orphans_since = existing.orphans_since.saturating_add(1); + existing } + None => StallReason { + // need to map out of `header` because they are not ser/de + block_id: header.map(|h| h.block_id), + block_hash: header.map(|h| h.hash), + prev_block_hash: header.map(|h| h.prev_block_hash), + first_seen: header.map(|h| h.timestamp), + l1_slot, + error, + orphans_since: 0, + }, + }; + self.set_stall_reason(&Some(stall)) + } - // Apply the clock invocation directly (it is expected to modify clock accounts). - let LeeTransaction::Public(clock_public_tx) = clock_tx else { - anyhow::bail!("Clock invocation must be a public transaction"); - }; - state_guard.transition_from_public_transaction( - clock_public_tx, - block.header.block_id, - block.header.timestamp, - )?; + /// Validates `block` against the tip and, if it chains, applies it atomically + /// (scratch clone, commit only on full success) and advances the tip. On any + /// failure records the stall and returns `Parked` without touching state. + pub async fn accept_block(&self, block: &Block, l1_slot: Slot) -> Result { + let tip = self.validated_tip()?; + + // Re-delivery of an already-applied block is idempotent, not a divergence + if let Some(tip) = &tip + && block.header.block_id <= tip.block_id + && let Some(stored) = self.get_block_at_id(block.header.block_id)? + && stored.header.hash == block.header.hash + { + return Ok(AcceptOutcome::AlreadyApplied); } - // ToDo: Currently we are fetching only finalized blocks - // if it changes, the following lines need to be updated - // to represent correct block finality - block.bedrock_status = BedrockStatus::Finalized; + if let Err(err) = chain_consistency::validate_against_tip(tip.as_ref(), block) { + self.record_stall(Some(&block.header), l1_slot, err.clone())?; + return Ok(AcceptOutcome::Parked(err)); + } - info!("Putting block {} into DB", block.header.block_id); - Ok(self.dbio.put_block(&block, l1_header.into())?) + // TODO: we use scratch state to be atomic, but need to revisit how expensive a clone is + let mut scratch = self.current_state.read().await.clone(); + if let Err(err) = chain_consistency::apply_block(block, &mut scratch) { + self.record_stall(Some(&block.header), l1_slot, err.clone())?; + return Ok(AcceptOutcome::Parked(err)); + } + + let mut stored = block.clone(); + stored.bedrock_status = BedrockStatus::Finalized; + self.dbio + .put_block(&stored, [0_u8; 32]) + .context("Failed to persist accepted block")?; + + // Commit in-memory state (infallible) only after the DB write succeeded. + *self.current_state.write().await = scratch; + // Best-effort: the block is durably applied, so a failed stall clear must not + // fail the apply. It self-heals on the next clear. + if let Err(err) = self.clear_stall_if_present() { + warn!("Failed to clear stall marker after applying block: {err:#}"); + } + Ok(AcceptOutcome::Applied) + } +} + +#[cfg(test)] +mod stall_reason_tests { + use common::HashType; + + use super::*; + use crate::stall_reason::StallReason; + + #[tokio::test] + async fn stall_reason_roundtrips_and_clears() { + let dir = tempfile::tempdir().expect("tempdir"); + let store = IndexerStore::open_db(dir.path()).expect("open store"); + + assert!(store.get_stall_reason().expect("get").is_none()); + + let stall = StallReason { + block_id: Some(7), + block_hash: Some(HashType([1_u8; 32])), + prev_block_hash: Some(HashType([2_u8; 32])), + l1_slot: Slot::from(42), + error: BlockIngestError::StateTransition { + tx_index: 0, + reason: "boom".to_owned(), + }, + first_seen: Some(99), + orphans_since: 3, + }; + store.set_stall_reason(&Some(stall)).expect("set stall"); + + let got = store.get_stall_reason().expect("get").expect("present"); + assert_eq!(got.block_id, Some(7)); + assert_eq!(got.orphans_since, 3); + assert!(matches!( + got.error, + BlockIngestError::StateTransition { .. } + )); + assert_eq!(got.block_hash, Some(HashType([1_u8; 32]))); + assert_eq!(got.prev_block_hash, Some(HashType([2_u8; 32]))); + assert_eq!(got.l1_slot, Slot::from(42)); + assert_eq!(got.first_seen, Some(99)); + + store.set_stall_reason(&None).expect("clear"); + assert!(store.get_stall_reason().expect("get").is_none()); } } #[cfg(test)] mod tests { - use common::{HashType, block::HashableBlockData}; + use common::test_utils::{create_transaction_native_token_transfer, produce_dummy_block}; use tempfile::tempdir; use testnet_initial_state::initial_pub_accounts_private_keys; use super::*; - struct TestFixture { - storage: IndexerStore, - from: AccountId, - to: AccountId, - _home: tempfile::TempDir, - } - - #[expect( - clippy::arithmetic_side_effects, - reason = "test helper with bounded inputs" - )] - async fn store_with_transfer_blocks( - block_count: u64, - prev_hash: Option, - ) -> TestFixture { - let home = tempdir().unwrap(); - let storage = IndexerStore::open_db(home.path()).unwrap(); - - let initial_accounts = initial_pub_accounts_private_keys(); - let from = initial_accounts[0].account_id; - let to = initial_accounts[1].account_id; - let sign_key = initial_accounts[0].pub_sign_key.clone(); - - let mut prev_hash = prev_hash; - for i in 0..block_count { - let tx = common::test_utils::create_transaction_native_token_transfer( - from, - u128::from(i), - to, - 10, - &sign_key, - ); - let block_id = i + 1; - - let next_block = common::test_utils::produce_dummy_block(block_id, prev_hash, vec![tx]); - prev_hash = Some(next_block.header.hash); - - storage - .put_block( - next_block, - HeaderId::from([u8::try_from(i + 1).unwrap(); 32]), - ) - .await - .unwrap(); - } - - TestFixture { - storage, - from, - to, - _home: home, - } - } - #[test] fn correct_startup() { let home = tempdir().unwrap(); @@ -282,75 +328,348 @@ mod tests { } #[tokio::test] - async fn state_transition() { + async fn accept_block_applies_transfers_and_advances_tip() { let home = tempdir().unwrap(); - let storage = IndexerStore::open_db(home.as_ref()).unwrap(); + let store = IndexerStore::open_db(home.as_ref()).unwrap(); let initial_accounts = initial_pub_accounts_private_keys(); let from = initial_accounts[0].account_id; let to = initial_accounts[1].account_id; let sign_key = initial_accounts[0].pub_sign_key.clone(); - let clock_tx = LeeTransaction::Public(clock_invocation(0)); - let genesis_block_data = HashableBlockData { - block_id: 1, - prev_block_hash: HashType::default(), - timestamp: 0, - transactions: vec![clock_tx], - }; - let genesis_block = genesis_block_data - .into_pending_block(&common::test_utils::sequencer_sign_key_for_testing()); - let mut prev_hash = Some(genesis_block.header.hash); - storage - .put_block(genesis_block, HeaderId::from([0_u8; 32])) - .await - .unwrap(); + // Genesis (block 1): clock-only. + let genesis = produce_dummy_block(1, None, vec![]); + let mut prev_hash = genesis.header.hash; + assert!(matches!( + store.accept_block(&genesis, Slot::from(0)).await.unwrap(), + AcceptOutcome::Applied + )); - for i in 0..10_u128 { - let tx = common::test_utils::create_transaction_native_token_transfer( - from, i, to, 10, &sign_key, - ); - let block_id = u64::try_from(i + 1).unwrap(); - let next_block = common::test_utils::produce_dummy_block(block_id, prev_hash, vec![tx]); - prev_hash = Some(next_block.header.hash); - storage - .put_block( - next_block, - HeaderId::from([u8::try_from(i + 1).unwrap(); 32]), - ) - .await - .unwrap(); + // Blocks 2..=11: one native transfer of 10 each (nonces 0..=9). + for i in 0..10_u64 { + let tx = create_transaction_native_token_transfer(from, i.into(), to, 10, &sign_key); + let block = produce_dummy_block(i + 2, Some(prev_hash), vec![tx]); + prev_hash = block.header.hash; + assert!(matches!( + store.accept_block(&block, Slot::from(0)).await.unwrap(), + AcceptOutcome::Applied + )); } - let acc1_val = storage.account_current_state(&from).await.unwrap(); - let acc2_val = storage.account_current_state(&to).await.unwrap(); - - assert_eq!(acc1_val.balance, 9900); - assert_eq!(acc2_val.balance, 20100); + assert_eq!( + store.account_current_state(&from).await.unwrap().balance, + 9900 + ); + assert_eq!( + store.account_current_state(&to).await.unwrap().balance, + 20100 + ); + // Tip advanced to the last applied block; a clean run leaves no stall. + assert_eq!(store.get_last_block_id().unwrap(), Some(11)); + assert!(store.get_stall_reason().unwrap().is_none()); } #[tokio::test] - async fn account_state_at_block() { - let TestFixture { - storage, - from, - to, - _home, - } = store_with_transfer_blocks(10, None).await; + async fn account_state_at_block_reflects_history() { + let home = tempdir().unwrap(); + let store = IndexerStore::open_db(home.as_ref()).unwrap(); - let acc1_at_1 = storage.account_state_at_block(&from, 1).unwrap(); - let acc2_at_1 = storage.account_state_at_block(&to, 1).unwrap(); - assert_eq!(acc1_at_1.balance, 9990); - assert_eq!(acc2_at_1.balance, 20010); + let initial_accounts = initial_pub_accounts_private_keys(); + let from = initial_accounts[0].account_id; + let to = initial_accounts[1].account_id; + let sign_key = initial_accounts[0].pub_sign_key.clone(); - let acc1_at_5 = storage.account_state_at_block(&from, 5).unwrap(); - let acc2_at_5 = storage.account_state_at_block(&to, 5).unwrap(); - assert_eq!(acc1_at_5.balance, 9950); - assert_eq!(acc2_at_5.balance, 20050); + let genesis = produce_dummy_block(1, None, vec![]); + let mut prev_hash = genesis.header.hash; + store.accept_block(&genesis, Slot::from(0)).await.unwrap(); - let acc1_at_9 = storage.account_state_at_block(&from, 9).unwrap(); - let acc2_at_9 = storage.account_state_at_block(&to, 9).unwrap(); - assert_eq!(acc1_at_9.balance, 9910); - assert_eq!(acc2_at_9.balance, 20090); + for i in 0..10_u64 { + let tx = create_transaction_native_token_transfer(from, i.into(), to, 10, &sign_key); + let block = produce_dummy_block(i + 2, Some(prev_hash), vec![tx]); + prev_hash = block.header.hash; + store.accept_block(&block, Slot::from(0)).await.unwrap(); + } + + // State at block N is inclusive of block N. + // Block 1 (genesis, clock-only): no transfers yet. + assert_eq!( + store.account_state_at_block(&from, 1).unwrap().balance, + 10000 + ); + assert_eq!(store.account_state_at_block(&to, 1).unwrap().balance, 20000); + // Through block 5: 4 transfers applied (blocks 2..=5). + assert_eq!( + store.account_state_at_block(&from, 5).unwrap().balance, + 9960 + ); + assert_eq!(store.account_state_at_block(&to, 5).unwrap().balance, 20040); + // Through block 9: 8 transfers applied (blocks 2..=9). + assert_eq!( + store.account_state_at_block(&from, 9).unwrap().balance, + 9920 + ); + assert_eq!(store.account_state_at_block(&to, 9).unwrap().balance, 20080); + } +} + +#[cfg(test)] +mod accept_tests { + use chain_consistency::BlockIngestError; + use common::{HashType, block::HashableBlockData, test_utils::produce_dummy_block}; + + use super::*; + + fn signing_key() -> lee::PrivateKey { + lee::PrivateKey::try_new([7_u8; 32]).expect("valid key") + } + + // A block with a correct hash but empty body — enough to exercise the + // acceptance checks (id/link/hash), which run before any state application. + fn valid_hash_block(block_id: u64, prev: HashType) -> common::block::Block { + HashableBlockData { + block_id, + prev_block_hash: prev, + timestamp: 0, + transactions: vec![], + } + .into_pending_block(&signing_key()) + } + + #[tokio::test] + async fn non_genesis_first_block_parks_with_unexpected_id() { + let dir = tempfile::tempdir().expect("tempdir"); + let store = IndexerStore::open_db(dir.path()).expect("open store"); + + let block = valid_hash_block(2, HashType([0_u8; 32])); + let outcome = store + .accept_block(&block, Slot::from(0)) + .await + .expect("accept"); + + assert!(matches!( + outcome, + AcceptOutcome::Parked(BlockIngestError::UnexpectedBlockId { + expected: 1, + got: 2 + }) + )); + let stall = store.get_stall_reason().expect("get").expect("present"); + assert_eq!(stall.block_id, Some(2)); + assert_eq!(stall.orphans_since, 0); + } + + #[tokio::test] + async fn hash_mismatch_parks() { + let dir = tempfile::tempdir().expect("tempdir"); + let store = IndexerStore::open_db(dir.path()).expect("open store"); + + let mut block = valid_hash_block(1, HashType([0_u8; 32])); + block.header.timestamp = 999; // invalidates the stored hash + + let outcome = store + .accept_block(&block, Slot::from(0)) + .await + .expect("accept"); + assert!(matches!( + outcome, + AcceptOutcome::Parked(BlockIngestError::HashMismatch { .. }) + )); + } + + #[tokio::test] + async fn second_break_bumps_orphan_count_and_keeps_first() { + let dir = tempfile::tempdir().expect("tempdir"); + let store = IndexerStore::open_db(dir.path()).expect("open store"); + + let first = valid_hash_block(2, HashType([0_u8; 32])); + store + .accept_block(&first, Slot::from(0)) + .await + .expect("accept"); + let second = valid_hash_block(3, HashType([0_u8; 32])); + store + .accept_block(&second, Slot::from(0)) + .await + .expect("accept"); + + let stall = store.get_stall_reason().expect("get").expect("present"); + assert_eq!(stall.block_id, Some(2), "first stall preserved"); + assert_eq!(stall.orphans_since, 1, "second break counted as orphan"); + } + + #[tokio::test] + async fn deserialize_break_records_stall_without_header() { + let dir = tempfile::tempdir().expect("tempdir"); + let store = IndexerStore::open_db(dir.path()).expect("open store"); + + store + .record_stall( + None, + Slot::from(0), + BlockIngestError::Deserialize("bad bytes".to_owned()), + ) + .expect("record"); + + let stall = store.get_stall_reason().expect("get").expect("present"); + assert_eq!(stall.block_id, None); + assert!(matches!(stall.error, BlockIngestError::Deserialize(_))); + } + + #[tokio::test] + async fn parks_then_recovers_on_valid_continuation() { + let dir = tempfile::tempdir().expect("tempdir"); + let store = IndexerStore::open_db(dir.path()).expect("open store"); + + // Genesis (block 1, clock-only) applies and advances the tip. + let genesis = produce_dummy_block(1, None, vec![]); + assert!(matches!( + store.accept_block(&genesis, Slot::from(0)).await.unwrap(), + AcceptOutcome::Applied + )); + + // A block that skips ahead (id 3 while the tip is 1) parks the indexer. + let bad = produce_dummy_block(3, Some(genesis.header.hash), vec![]); + assert!(matches!( + store.accept_block(&bad, Slot::from(0)).await.unwrap(), + AcceptOutcome::Parked(BlockIngestError::UnexpectedBlockId { + expected: 2, + got: 3 + }) + )); + assert!( + store.get_stall_reason().unwrap().is_some(), + "indexer should be parked after the bad block" + ); + assert_eq!( + store.get_last_block_id().unwrap(), + Some(1), + "validated tip must stay frozen at genesis while parked" + ); + + // The valid continuation (block 2 chaining on genesis) recovers the chain. + let next = produce_dummy_block(2, Some(genesis.header.hash), vec![]); + assert!(matches!( + store.accept_block(&next, Slot::from(0)).await.unwrap(), + AcceptOutcome::Applied + )); + assert!( + store.get_stall_reason().unwrap().is_none(), + "stall reason must clear on recovery" + ); + assert_eq!( + store.get_last_block_id().unwrap(), + Some(2), + "tip must advance to the recovered block" + ); + } + + #[tokio::test] + async fn redelivered_tip_block_is_idempotent_not_parked() { + use testnet_initial_state::initial_pub_accounts_private_keys; + + let dir = tempfile::tempdir().expect("tempdir"); + let store = IndexerStore::open_db(dir.path()).expect("open store"); + + let accounts = initial_pub_accounts_private_keys(); + let from = accounts[0].account_id; + let to = accounts[1].account_id; + let sign_key = accounts[0].pub_sign_key.clone(); + + let genesis = produce_dummy_block(1, None, vec![]); + store + .accept_block(&genesis, Slot::from(0)) + .await + .expect("accept genesis"); + + // Block 2: a single transfer of 10. + let tx = common::test_utils::create_transaction_native_token_transfer( + from, 0, to, 10, &sign_key, + ); + let block = produce_dummy_block(2, Some(genesis.header.hash), vec![tx]); + assert!(matches!( + store.accept_block(&block, Slot::from(0)).await.unwrap(), + AcceptOutcome::Applied + )); + let balance_after = store.account_current_state(&from).await.unwrap().balance; + + // Re-deliver the exact same block: idempotent skip, no state change, no park. + assert!(matches!( + store.accept_block(&block, Slot::from(0)).await.unwrap(), + AcceptOutcome::AlreadyApplied + )); + assert_eq!( + store.account_current_state(&from).await.unwrap().balance, + balance_after, + "re-delivered block must not be applied twice" + ); + assert_eq!( + store.get_last_block_id().unwrap(), + Some(2), + "tip must stay at the already-applied block" + ); + assert!( + store.get_stall_reason().unwrap().is_none(), + "a benign duplicate must not park the indexer" + ); + } + + #[tokio::test] + async fn redelivered_block_below_tip_is_idempotent_not_parked() { + use testnet_initial_state::initial_pub_accounts_private_keys; + + let dir = tempfile::tempdir().expect("tempdir"); + let store = IndexerStore::open_db(dir.path()).expect("open store"); + + let accounts = initial_pub_accounts_private_keys(); + let from = accounts[0].account_id; + let to = accounts[1].account_id; + let sign_key = accounts[0].pub_sign_key.clone(); + + // Build a short chain: genesis (1) -> block 2 -> block 3, so the tip is 3. + let genesis = produce_dummy_block(1, None, vec![]); + store + .accept_block(&genesis, Slot::from(0)) + .await + .expect("accept genesis"); + + let tx2 = common::test_utils::create_transaction_native_token_transfer( + from, 0, to, 10, &sign_key, + ); + let block2 = produce_dummy_block(2, Some(genesis.header.hash), vec![tx2]); + assert!(matches!( + store.accept_block(&block2, Slot::from(0)).await.unwrap(), + AcceptOutcome::Applied + )); + + let tx3 = common::test_utils::create_transaction_native_token_transfer( + from, 1, to, 10, &sign_key, + ); + let block3 = produce_dummy_block(3, Some(block2.header.hash), vec![tx3]); + assert!(matches!( + store.accept_block(&block3, Slot::from(0)).await.unwrap(), + AcceptOutcome::Applied + )); + + let balance_after = store.account_current_state(&from).await.unwrap().balance; + + // Re-deliver block 2 (id below the tip): a re-delivery, not a divergence. + assert!(matches!( + store.accept_block(&block2, Slot::from(0)).await.unwrap(), + AcceptOutcome::AlreadyApplied + )); + assert_eq!( + store.account_current_state(&from).await.unwrap().balance, + balance_after, + "re-delivered block below the tip must not be applied again" + ); + assert_eq!( + store.get_last_block_id().unwrap(), + Some(3), + "tip must stay at the current head" + ); + assert!( + store.get_stall_reason().unwrap().is_none(), + "a benign re-delivery must not park the indexer" + ); } } diff --git a/lez/indexer/core/src/config.rs b/lez/indexer/core/src/config.rs index cb7f3dfe..c069d5cf 100644 --- a/lez/indexer/core/src/config.rs +++ b/lez/indexer/core/src/config.rs @@ -20,6 +20,13 @@ pub struct IndexerConfig { pub consensus_info_polling_interval: Duration, pub bedrock_config: ClientConfig, pub channel_id: ChannelId, + /// Whether to wipe the indexer store and re-index from scratch when the startup + /// chain-identity check finds the channel serving a different block than the one + /// stored at the same id. + /// + /// Defaults to `false`: on mismatch the indexer refuses to start. + #[serde(default)] + pub allow_chain_reset: bool, } impl IndexerConfig { diff --git a/lez/indexer/core/src/lib.rs b/lez/indexer/core/src/lib.rs index 0d595fc0..04247140 100644 --- a/lez/indexer/core/src/lib.rs +++ b/lez/indexer/core/src/lib.rs @@ -2,28 +2,31 @@ use std::{path::Path, sync::Arc}; use anyhow::Result; use arc_swap::ArcSwap; +use chain_consistency::{Anchor, BlockIngestError, ChainConsistency}; use common::block::Block; -// ToDo: Remove after testnet +// TODO: Remove after testnet use futures::StreamExt as _; use log::{error, info, warn}; -use logos_blockchain_core::header::HeaderId; use logos_blockchain_zone_sdk::{ - CommonHttpClient, ZoneMessage, adapter::NodeHttpClient, indexer::ZoneIndexer, + CommonHttpClient, Slot, ZoneMessage, adapter::NodeHttpClient, indexer::ZoneIndexer, }; +pub use stall_reason::StallReason; use crate::{ - block_store::IndexerStore, + block_store::{AcceptOutcome, IndexerStore}, config::IndexerConfig, status::{IndexerStatus, IndexerSyncStatus}, }; - pub mod block_store; pub mod config; +pub mod stall_reason; pub mod status; #[derive(Clone)] pub struct IndexerCore { pub zone_indexer: Arc>, + /// Direct node handle for queries outside `ZoneIndexer`'s streaming API. + pub node: NodeHttpClient, pub config: IndexerConfig, pub store: IndexerStore, /// Live ingestion status; updated by the ingest stream, read by `status`. @@ -31,7 +34,42 @@ pub struct IndexerCore { } impl IndexerCore { - pub fn new(config: IndexerConfig, storage_dir: &Path) -> Result { + /// Builds the core, then verifies the stored chain matches the channel's by + /// re-reading the channel at the stored tip's position. + /// + /// On mismatch: refuse (error) unless `config.allow_chain_reset` is set, in which case wipe the + /// store and re-index from scratch. + pub async fn new(config: IndexerConfig, storage_dir: &Path) -> Result { + let home = storage_dir.join(format!("rocksdb-{}", config.channel_id)); + let core = Self::open(config.clone(), storage_dir)?; + match core.verify_chain_consistency().await? { + // `Inconclusive` is deliberately treated the same as `Consistent`. + // + // We could not prove a reset, so proceed from the cursor without wiping + // a possibly-valid store. A genuinely divergent chain is still caught + // later when the ingest loop tries to apply and parks. + ChainConsistency::Consistent | ChainConsistency::Inconclusive => Ok(core), + ChainConsistency::Inconsistent(mismatch) if config.allow_chain_reset => { + warn!( + "Chain reset detected ({mismatch}). Wiping indexer store at {} and \ + re-indexing.", + home.display() + ); + drop(core); // sole owner before the ingest task is spawned → closes the DB + storage::indexer::RocksDBIO::destroy(&home)?; + Self::open(config, storage_dir) + } + ChainConsistency::Inconsistent(mismatch) => Err(anyhow::anyhow!( + "Indexer store at {} holds a different chain than the channel now serves \ + ({mismatch}). Delete the indexer storage directory, point at a fresh one, or \ + set `allow_chain_reset` in the indexer config.", + home.display() + )), + } + } + + /// Opens the store and builds the core without the chain-identity check. + fn open(config: IndexerConfig, storage_dir: &Path) -> Result { // Namespace the DB by channel so indexers on different channels can // share a storage dir without their RocksDB state colliding. let home = storage_dir.join(format!("rocksdb-{}", config.channel_id)); @@ -41,16 +79,46 @@ impl IndexerCore { CommonHttpClient::new(basic_auth), config.bedrock_config.addr.clone(), ); - let zone_indexer = ZoneIndexer::new(config.channel_id, node); + let zone_indexer = ZoneIndexer::new(config.channel_id, node.clone()); Ok(Self { zone_indexer: Arc::new(zone_indexer), + node, config, store: IndexerStore::open_db(&home)?, status: Arc::new(ArcSwap::from_pointee(IndexerSyncStatus::starting())), }) } + /// Detects when the local store belongs to a different chain than the connected + /// L1 (e.g. a wiped/restarted bedrock) so startup can reset instead of silently + /// diverging. Best-effort dev convenience; won't trigger during normal live indexing. + /// + /// The anchoring/probe logic lives in the shared [`chain_consistency`] crate; + /// here we build the [`Anchor`] from the store (the recorded parked block while + /// stalled, otherwise the tip at the read cursor — only blocks advance the + /// cursor, so when not parked it is the tip's inscription slot) and delegate. + async fn verify_chain_consistency(&self) -> Result { + let anchor = if let Some(stall) = self.store.get_stall_reason()? { + Anchor::new(stall.l1_slot, stall.block_id.zip(stall.block_hash)) + } else { + let Some(cursor) = self.store.get_zone_cursor()? else { + // if we have no cursor, the store is empty or cold: nothing to compare + return Ok(ChainConsistency::Inconclusive); + }; + let Some(tip_id) = self.store.get_last_block_id()? else { + return Ok(ChainConsistency::Inconclusive); + }; + let Some(tip) = self.store.get_block_at_id(tip_id)? else { + return Ok(ChainConsistency::Inconclusive); + }; + Anchor::new(cursor, Some((tip_id, tip.header.hash))) + }; + + chain_consistency::verify_chain_consistency(&self.node, self.config.channel_id, &anchor) + .await + } + /// Snapshot of the current ingestion status (sync state + indexed tip). /// /// Combines the ingest loop's live status with the L2 tip read fresh from the @@ -58,10 +126,27 @@ impl IndexerCore { #[must_use] pub fn status(&self) -> IndexerStatus { let sync = IndexerSyncStatus::clone(&self.status.load()); - let indexed_block_id = self.store.get_last_block_id().ok().flatten(); + // Log-and-fall-back rather than collapsing a store error into the same + // `None` as "legitimately absent": a DB read failure must not silently + // masquerade as "no tip yet" / "no stall recorded" in the snapshot. + let indexed_block_id = match self.store.get_last_block_id() { + Ok(id) => id, + Err(err) => { + warn!("Failed to read last indexed block id for status: {err:#}"); + None + } + }; + let stall_reason = match self.store.get_stall_reason() { + Ok(reason) => reason, + Err(err) => { + warn!("Failed to read stall reason for status: {err:#}"); + None + } + }; IndexerStatus { sync, indexed_block_id, + stall_reason, } } @@ -70,6 +155,35 @@ impl IndexerCore { self.status.store(Arc::new(status)); } + /// Advances the in-memory L1 read cursor past `slot` and persists it. + /// A persist failure is only logged: the worst case is re-reading a batch + /// after a restart, which ingestion handles idempotently. + fn advance_cursor(&self, cursor: &mut Option, slot: Slot) { + *cursor = Some(slot); + if let Err(err) = self.store.set_zone_cursor(&slot) { + warn!("Failed to persist indexer cursor: {err:#}"); + } + } + + /// Parks on an inscription that could not be parsed as an L2 block: + /// records the stall and flips the status. The validated tip stays frozen. + fn park_undeserializable(&self, slot: Slot, error: std::io::Error) { + let error = anyhow::Error::new(error); + + // use `:#` to get the entire error chain + let reason = format!("{error:#}"); + error!("Failed to deserialize L2 block from zone-sdk: {reason}"); + if let Err(err) = + self.store + .record_stall(None, slot, BlockIngestError::Deserialize(reason.clone())) + { + warn!("Failed to record stall reason: {err:#}"); + } + self.set_status(IndexerSyncStatus::stalled(format!( + "failed to deserialize L2 block: {reason}" + ))); + } + pub fn subscribe_parse_block_stream(&self) -> impl futures::Stream> + '_ { let poll_interval = self.config.consensus_info_polling_interval; let initial_cursor = self @@ -80,8 +194,8 @@ impl IndexerCore { async_stream::stream! { let mut cursor = initial_cursor; - if cursor.is_some() { - info!("Resuming indexer from cursor {cursor:?}"); + if let Some(slot) = &cursor { + info!("Resuming indexer from cursor {slot:?}"); } else { info!("Starting indexer from beginning of channel"); } @@ -90,8 +204,6 @@ impl IndexerCore { let stream = match self.zone_indexer.next_messages(cursor).await { Ok(s) => s, Err(err) => { - // `next_messages` reads L1 consensus info internally, so - // this also covers an unreachable/misconfigured L1 node. error!("Failed to start zone-sdk next_messages stream: {err}"); self.set_status(IndexerSyncStatus::error(format!( "cannot reach L1 / read channel: {err}" @@ -102,11 +214,8 @@ impl IndexerCore { }; let mut stream = std::pin::pin!(stream); - // Flip to Syncing on the first message of this cycle (not merely on - // a successful poll) so the steady-state CaughtUp status doesn't - // flicker. Until then the state stays Starting (cold-start scan of - // empty L1 history) or CaughtUp (idle). let mut announced_syncing = false; + let mut had_cycle_error = false; while let Some((msg, slot)) = stream.next().await { if !announced_syncing { @@ -116,48 +225,163 @@ impl IndexerCore { let zone_block = match msg { ZoneMessage::Block(b) => b, - // Non-block messages don't carry a cursor position; the - // next ZoneBlock advances past them implicitly. + // FIXME: will be handled in prep of decentralized sequencers ZoneMessage::Deposit(_) | ZoneMessage::Withdraw(_) => continue, }; let block: Block = match borsh::from_slice(&zone_block.data) { Ok(b) => b, - Err(e) => { - error!("Failed to deserialize L2 block from zone-sdk: {e}"); - // Advance past the broken inscription so we don't - // re-process it on restart. - cursor = Some(slot); - if let Err(err) = self.store.set_zone_cursor(&slot) { - warn!("Failed to persist indexer cursor: {err:#}"); - } + Err(error) => { + self.park_undeserializable(slot, error); + // L1 proceeds regardless + self.advance_cursor(&mut cursor, slot); continue; } }; - info!("Indexed L2 block {}", block.header.block_id); - - // TODO: Remove l1_header placeholder once storage layer - // no longer requires it. Zone-sdk handles L1 tracking internally. - let placeholder_l1_header = HeaderId::from([0_u8; 32]); - if let Err(err) = self.store.put_block(block.clone(), placeholder_l1_header).await { - error!("Failed to store block {}: {err:#}", block.header.block_id); + match self.store.accept_block(&block, slot).await { + Ok(AcceptOutcome::Applied) => { + info!("Indexed L2 block {}", block.header.block_id); + self.set_status(IndexerSyncStatus::syncing()); + self.advance_cursor(&mut cursor, slot); + yield Ok(block); + } + Ok(AcceptOutcome::AlreadyApplied) => { + info!( + "Skipping already-applied block {}", + block.header.block_id + ); + self.advance_cursor(&mut cursor, slot); + } + Ok(AcceptOutcome::Parked(ingest_err)) => { + error!( + "Parked at block {}: {ingest_err}", + block.header.block_id + ); + self.set_status(IndexerSyncStatus::stalled(ingest_err.to_string())); + // L1 proceeds regardless + self.advance_cursor(&mut cursor, slot); + } + Err(err) => { + // Infrastructure error (DB read/write), not a bad block. + // will re-poll from the same cursor next cycle. + error!( + "Store error applying block {}: {err:#}", + block.header.block_id + ); + self.set_status(IndexerSyncStatus::error(format!( + "store error: {err:#}" + ))); + had_cycle_error = true; + break; + } } - - cursor = Some(slot); - if let Err(err) = self.store.set_zone_cursor(&slot) { - warn!("Failed to persist indexer cursor: {err:#}"); - } - yield Ok(block); } - // Stream drained: caught up to LIB as of this cycle. Clears any - // prior error (e.g. a transient L1 disconnect that left no - // backlog, so the `Syncing` branch above never ran). Sleep then - // poll again. - self.set_status(IndexerSyncStatus::caught_up()); + if had_cycle_error { + tokio::time::sleep(poll_interval).await; + continue; + } + + // Stream drained. Stay Stalled if parked; otherwise we are caught up. + // A store error here must not be collapsed to "no stall recorded": + // that would wrongly flip us to caught-up, so we log and hold state. + match self.store.get_stall_reason() { + Ok(None) => self.set_status(IndexerSyncStatus::caught_up()), + Ok(Some(_)) => {} + Err(err) => { + warn!("Failed to read stall reason after draining stream; not marking caught up: {err:#}"); + } + } tokio::time::sleep(poll_interval).await; } } } } + +#[cfg(test)] +mod tests { + use std::time::Duration; + + use common::{HashType, block::HashableBlockData}; + use logos_blockchain_zone_sdk::Slot; + + use super::*; + use crate::config::{ChannelId, ClientConfig, IndexerConfig}; + + fn unreachable_core(dir: &std::path::Path) -> IndexerCore { + let config = IndexerConfig { + consensus_info_polling_interval: Duration::from_secs(1), + bedrock_config: ClientConfig { + addr: "http://localhost:1".parse().expect("url"), + auth: None, + }, + channel_id: ChannelId::from([1; 32]), + allow_chain_reset: false, + }; + IndexerCore::open(config, dir).expect("open core") + } + + fn test_block(block_id: u64, timestamp: u64) -> Block { + HashableBlockData { + block_id, + prev_block_hash: HashType([0; 32]), + timestamp, + transactions: vec![], + } + .into_pending_block(&lee::PrivateKey::try_new([7; 32]).expect("valid key")) + } + + #[tokio::test] + async fn cold_store_is_inconclusive() { + // An empty store has no cursor, so there is nothing to compare: the check + // must be Inconclusive (not Consistent), and it returns before any L1 read. + let dir = tempfile::tempdir().expect("tempdir"); + let core = unreachable_core(dir.path()); + assert!(matches!( + core.verify_chain_consistency().await.expect("verify"), + ChainConsistency::Inconclusive + )); + } + + #[tokio::test] + async fn parked_store_with_unreachable_node_is_inconclusive() { + // Network failure is not evidence of a reset: a parked store must stay + // parked (Inconclusive), not error out or trip the wipe path. + let dir = tempfile::tempdir().expect("tempdir"); + let core = unreachable_core(dir.path()); + let parked = test_block(5, 42); + core.store + .record_stall( + Some(&parked.header), + Slot::from(1_000), + BlockIngestError::EmptyBlock, + ) + .expect("record stall"); + assert!(matches!( + core.verify_chain_consistency().await.expect("verify"), + ChainConsistency::Inconclusive + )); + } + + #[tokio::test] + async fn caught_up_store_with_unreachable_node_is_inconclusive() { + let dir = tempfile::tempdir().expect("tempdir"); + let core = unreachable_core(dir.path()); + let genesis = common::test_utils::produce_dummy_block(1, None, vec![]); + assert!(matches!( + core.store + .accept_block(&genesis, Slot::from(1_000)) + .await + .expect("accept"), + AcceptOutcome::Applied + )); + core.store + .set_zone_cursor(&Slot::from(1_000)) + .expect("set cursor"); + assert!(matches!( + core.verify_chain_consistency().await.expect("verify"), + ChainConsistency::Inconclusive + )); + } +} diff --git a/lez/indexer/core/src/stall_reason.rs b/lez/indexer/core/src/stall_reason.rs new file mode 100644 index 00000000..c5118fcc --- /dev/null +++ b/lez/indexer/core/src/stall_reason.rs @@ -0,0 +1,25 @@ +use common::HashType; +use logos_blockchain_zone_sdk::Slot; +use serde::{Deserialize, Serialize}; + +use crate::ingest_error::BlockIngestError; + +/// Diagnostic record of the first block that broke the L2 chain. +/// +/// The block-derived fields are `None` for a deserialize break (no header was +/// ever parsed). `l1_slot` is the L1 slot the breaking inscription was read at. +/// `first_seen` is the breaking block's L2 timestamp (`None` for a deserialize break). +#[derive(Debug, Clone, Serialize, Deserialize)] +pub struct StallReason { + pub block_id: Option, + pub block_hash: Option, + pub prev_block_hash: Option, + pub l1_slot: Slot, + pub error: BlockIngestError, + pub first_seen: Option, + /// Number of later non-chaining blocks (orphans, since the tip is frozen). + /// + /// TODO: We could store a different "branch" of blocks following this break, but for now we + /// just count them. + pub orphans_since: u64, +} diff --git a/lez/indexer/core/src/status.rs b/lez/indexer/core/src/status.rs index 1193e124..a483fde6 100644 --- a/lez/indexer/core/src/status.rs +++ b/lez/indexer/core/src/status.rs @@ -1,9 +1,10 @@ use serde::Serialize; +use crate::stall_reason::StallReason; + /// Coarse lifecycle state of the indexer's ingestion loop, so a client can tell /// "still catching up" apart from "something went wrong". #[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize)] -#[serde(rename_all = "snake_case")] pub enum IndexerSyncState { /// Booted; no ingestion cycle has run yet. Starting, @@ -13,12 +14,14 @@ pub enum IndexerSyncState { CaughtUp, /// The last cycle failed (e.g. the L1 node is unreachable). See `last_error`. Error, + /// Parked on a stall reason: the validated tip is frozen awaiting a valid + /// continuation. See `last_error` and the snapshot's `stall_reason`. + Stalled, } /// Live ingestion status owned by the ingest loop: the coarse `state` plus the /// reason when it is `Error`. #[derive(Debug, Clone, Serialize)] -#[serde(rename_all = "camelCase")] pub struct IndexerSyncStatus { pub state: IndexerSyncState, pub last_error: Option, @@ -56,6 +59,15 @@ impl IndexerSyncStatus { last_error: Some(reason), } } + + /// Parked on a stall reason; `reason` mirrors the stall's error message. + /// The full stall is attached to the [`IndexerStatus`] snapshot. + pub(crate) const fn stalled(reason: String) -> Self { + Self { + state: IndexerSyncState::Stalled, + last_error: Some(reason), + } + } } /// Full status snapshot returned to callers (FFI/RPC): the live [`IndexerSyncStatus`] @@ -64,11 +76,11 @@ impl IndexerSyncStatus { /// The tip is tracked by the store, not the ingest loop, so it lives here on the /// returned snapshot rather than inside the shared [`IndexerSyncStatus`]. #[derive(Debug, Clone, Serialize)] -#[serde(rename_all = "camelCase")] pub struct IndexerStatus { #[serde(flatten)] pub sync: IndexerSyncStatus, pub indexed_block_id: Option, + pub stall_reason: Option, } #[cfg(test)] @@ -80,14 +92,16 @@ mod tests { let status = IndexerStatus { sync: IndexerSyncStatus::error("boom".to_owned()), indexed_block_id: Some(7), + stall_reason: None, }; let value = serde_json::to_value(&status).expect("serialize"); assert_eq!( value, serde_json::json!({ - "state": "error", - "lastError": "boom", - "indexedBlockId": 7, + "state": "Error", + "last_error": "boom", + "indexed_block_id": 7, + "stall_reason": null, }) ); } @@ -97,7 +111,36 @@ mod tests { let value = serde_json::to_value(IndexerSyncStatus::caught_up()).expect("serialize"); assert_eq!( value, - serde_json::json!({ "state": "caught_up", "lastError": null }) + serde_json::json!({ "state": "CaughtUp", "last_error": null }) ); } + + #[test] + fn stalled_status_serializes_with_stall_reason() { + use logos_blockchain_zone_sdk::Slot; + + use crate::{ingest_error::BlockIngestError, stall_reason::StallReason}; + + let status = IndexerStatus { + sync: IndexerSyncStatus::stalled("broken chain link".to_owned()), + indexed_block_id: Some(41), + stall_reason: Some(StallReason { + block_id: Some(42), + block_hash: None, + prev_block_hash: None, + l1_slot: Slot::from(0), + error: BlockIngestError::StateTransition { + tx_index: 0, + reason: String::default(), + }, + first_seen: None, + orphans_since: 2, + }), + }; + let value = serde_json::to_value(&status).expect("serialize"); + assert_eq!(value["state"], serde_json::json!("Stalled")); + assert_eq!(value["last_error"], serde_json::json!("broken chain link")); + assert_eq!(value["indexed_block_id"], serde_json::json!(41)); + assert_eq!(value["stall_reason"]["orphans_since"], serde_json::json!(2)); + } } diff --git a/lez/indexer/ffi/indexer_ffi.h b/lez/indexer/ffi/indexer_ffi.h index 8347ad3c..26857af7 100644 --- a/lez/indexer/ffi/indexer_ffi.h +++ b/lez/indexer/ffi/indexer_ffi.h @@ -503,9 +503,9 @@ struct LastBlockIdResult query_last_block(const struct IndexerServiceFFI *indexe * Query the indexer's current sync status as a JSON C-string. * * The JSON schema is owned by `indexer_core` (`IndexerStatus`): an object with - * `state` (`starting`/`syncing`/`caught_up`/`error`), `indexedBlockId`, and - * `lastError`. Lets a client distinguish "still catching up" from "something - * went wrong". + * `state` (`Starting`/`Syncing`/`CaughtUp`/`Error`/`Stalled`), + * `indexed_block_id`, `last_error`, and `stall_reason`. Lets a client + * distinguish "still catching up" from "something went wrong". * * # Arguments * diff --git a/lez/indexer/ffi/src/api/lifecycle.rs b/lez/indexer/ffi/src/api/lifecycle.rs index f668f3ee..8a1eafaf 100644 --- a/lez/indexer/ffi/src/api/lifecycle.rs +++ b/lez/indexer/ffi/src/api/lifecycle.rs @@ -112,10 +112,12 @@ unsafe fn setup_indexer( unsafe { Runtime::from_borrowed(caller.as_ref()) } }; - let core = IndexerCore::new(config, &storage_dir).map_err(|e| { - log::error!("Could not initialize indexer core: {e}"); - OperationStatus::InitializationError - })?; + let core = runtime + .block_on(IndexerCore::new(config, &storage_dir)) + .map_err(|e| { + log::error!("Could not initialize indexer core: {e}"); + OperationStatus::InitializationError + })?; // The block stream writes each parsed block into the store as a side effect // of being polled, so we spawn a task that simply drains it. There are no diff --git a/lez/indexer/ffi/src/api/query.rs b/lez/indexer/ffi/src/api/query.rs index 1943f6d4..dff027fa 100644 --- a/lez/indexer/ffi/src/api/query.rs +++ b/lez/indexer/ffi/src/api/query.rs @@ -91,9 +91,9 @@ pub unsafe extern "C" fn query_last_block(indexer: *const IndexerServiceFFI) -> /// Query the indexer's current sync status as a JSON C-string. /// /// The JSON schema is owned by `indexer_core` (`IndexerStatus`): an object with -/// `state` (`starting`/`syncing`/`caught_up`/`error`), `indexedBlockId`, and -/// `lastError`. Lets a client distinguish "still catching up" from "something -/// went wrong". +/// `state` (`Starting`/`Syncing`/`CaughtUp`/`Error`/`Stalled`), +/// `indexed_block_id`, `last_error`, and `stall_reason`. Lets a client +/// distinguish "still catching up" from "something went wrong". /// /// # Arguments /// diff --git a/lez/indexer/service/configs/debug/indexer_config.json b/lez/indexer/service/configs/debug/indexer_config.json index 85227700..be5cf353 100644 --- a/lez/indexer/service/configs/debug/indexer_config.json +++ b/lez/indexer/service/configs/debug/indexer_config.json @@ -1,7 +1,8 @@ { - "consensus_info_polling_interval": "1s", - "bedrock_config": { - "addr": "http://localhost:18080" - }, - "channel_id": "0101010101010101010101010101010101010101010101010101010101010101" + "consensus_info_polling_interval": "1s", + "bedrock_config": { + "addr": "http://localhost:18080" + }, + "channel_id": "0101010101010101010101010101010101010101010101010101010101010101", + "allow_chain_reset": true } diff --git a/lez/indexer/service/configs/docker/indexer_config.json b/lez/indexer/service/configs/docker/indexer_config.json index f083ca27..ce28af0b 100644 --- a/lez/indexer/service/configs/docker/indexer_config.json +++ b/lez/indexer/service/configs/docker/indexer_config.json @@ -3,5 +3,6 @@ "bedrock_config": { "addr": "http://host.docker.internal:18080" }, - "channel_id": "0101010101010101010101010101010101010101010101010101010101010101" + "channel_id": "0101010101010101010101010101010101010101010101010101010101010101", + "allow_chain_reset": true } diff --git a/lez/indexer/service/protocol/Cargo.toml b/lez/indexer/service/protocol/Cargo.toml index 5a4176f5..b3e8f65c 100644 --- a/lez/indexer/service/protocol/Cargo.toml +++ b/lez/indexer/service/protocol/Cargo.toml @@ -11,6 +11,7 @@ workspace = true lee_core = { workspace = true, optional = true, features = ["host"] } lee = { workspace = true, optional = true } common = { workspace = true, optional = true } +indexer_core = { workspace = true, optional = true } serde = { workspace = true, features = ["derive"] } serde_with.workspace = true @@ -22,4 +23,4 @@ anyhow.workspace = true [features] # Enable conversion to/from LEE core types -convert = ["dep:lee_core", "dep:lee", "dep:common"] +convert = ["dep:lee_core", "dep:lee", "dep:common", "dep:indexer_core"] diff --git a/lez/indexer/service/protocol/src/convert.rs b/lez/indexer/service/protocol/src/convert.rs index cd0bff7e..55c4dc6c 100644 --- a/lez/indexer/service/protocol/src/convert.rs +++ b/lez/indexer/service/protocol/src/convert.rs @@ -3,11 +3,12 @@ use lee_core::account::Nonce; use crate::{ - Account, AccountId, BedrockStatus, Block, BlockBody, BlockHeader, Ciphertext, Commitment, - CommitmentSetDigest, Data, EncryptedAccountData, EphemeralPublicKey, HashType, Nullifier, - PrivacyPreservingMessage, PrivacyPreservingTransaction, ProgramDeploymentMessage, - ProgramDeploymentTransaction, ProgramId, Proof, PublicKey, PublicMessage, PublicTransaction, - Signature, Transaction, ValidityWindow, WitnessSet, + Account, AccountId, BedrockStatus, Block, BlockBody, BlockHeader, BlockIngestError, Ciphertext, + Commitment, CommitmentSetDigest, Data, EncryptedAccountData, EphemeralPublicKey, HashType, + IndexerStatus, IndexerSyncState, Nullifier, PrivacyPreservingMessage, + PrivacyPreservingTransaction, ProgramDeploymentMessage, ProgramDeploymentTransaction, + ProgramId, Proof, PublicKey, PublicMessage, PublicTransaction, Signature, StallReason, + Transaction, ValidityWindow, WitnessSet, }; // ============================================================================ @@ -707,3 +708,94 @@ impl TryFrom for lee_core::program::ValidityWindow { value.0.try_into() } } + +// ============================================================================ +// Indexer status conversions +// ============================================================================ + +impl From for IndexerSyncState { + fn from(value: indexer_core::status::IndexerSyncState) -> Self { + match value { + indexer_core::status::IndexerSyncState::Starting => Self::Starting, + indexer_core::status::IndexerSyncState::Syncing => Self::Syncing, + indexer_core::status::IndexerSyncState::CaughtUp => Self::CaughtUp, + indexer_core::status::IndexerSyncState::Error => Self::Error, + indexer_core::status::IndexerSyncState::Stalled => Self::Stalled, + } + } +} + +impl From for BlockIngestError { + fn from(value: indexer_core::BlockIngestError) -> Self { + match value { + indexer_core::BlockIngestError::Deserialize(msg) => Self::Deserialize(msg), + indexer_core::BlockIngestError::UnexpectedBlockId { expected, got } => { + Self::UnexpectedBlockId { expected, got } + } + indexer_core::BlockIngestError::BrokenChainLink { + expected_prev, + got_prev, + } => Self::BrokenChainLink { + expected_prev: expected_prev.into(), + got_prev: got_prev.into(), + }, + indexer_core::BlockIngestError::HashMismatch { computed, header } => { + Self::HashMismatch { + computed: computed.into(), + header: header.into(), + } + } + indexer_core::BlockIngestError::EmptyBlock => Self::EmptyBlock, + indexer_core::BlockIngestError::InvalidClockTransaction => { + Self::InvalidClockTransaction + } + indexer_core::BlockIngestError::NonPublicGenesisTransaction => { + Self::NonPublicGenesisTransaction + } + indexer_core::BlockIngestError::StateTransition { tx_index, reason } => { + Self::StateTransition { tx_index, reason } + } + } + } +} + +impl From for StallReason { + fn from(value: indexer_core::StallReason) -> Self { + let indexer_core::StallReason { + block_id, + block_hash, + prev_block_hash, + l1_slot, + error, + first_seen, + orphans_since, + } = value; + + Self { + block_id, + block_hash: block_hash.map(Into::into), + prev_block_hash: prev_block_hash.map(Into::into), + l1_slot: l1_slot.into_inner(), + error: error.into(), + first_seen, + orphans_since, + } + } +} + +impl From for IndexerStatus { + fn from(value: indexer_core::status::IndexerStatus) -> Self { + let indexer_core::status::IndexerStatus { + sync, + indexed_block_id, + stall_reason, + } = value; + + Self { + state: sync.state.into(), + last_error: sync.last_error, + indexed_block_id, + stall_reason: stall_reason.map(Into::into), + } + } +} diff --git a/lez/indexer/service/protocol/src/lib.rs b/lez/indexer/service/protocol/src/lib.rs index a670dee6..e17d539b 100644 --- a/lez/indexer/service/protocol/src/lib.rs +++ b/lez/indexer/service/protocol/src/lib.rs @@ -363,3 +363,73 @@ pub enum BedrockStatus { Safe, Finalized, } + +/// Coarse lifecycle state of the indexer's ingestion loop, so a client can tell +/// "still catching up" apart from "something went wrong". +#[derive(Debug, Copy, Clone, PartialEq, Eq, Hash, Serialize, Deserialize, JsonSchema)] +pub enum IndexerSyncState { + /// Booted; no ingestion cycle has run yet. + Starting, + /// Streaming finalized messages toward the L1 frontier. + Syncing, + /// Drained the stream up to the last finalized block; idle until new blocks finalize. + CaughtUp, + /// The last cycle failed (e.g. the L1 node is unreachable). See `last_error`. + Error, + /// Parked on a stall reason: the validated tip is frozen awaiting a valid + /// continuation. See `last_error` and `stall_reason`. + Stalled, +} + +/// Why the indexer could not apply an L2 block from the channel. +#[derive(Debug, Clone, PartialEq, Eq, Hash, Serialize, Deserialize, JsonSchema)] +pub enum BlockIngestError { + Deserialize(String), + UnexpectedBlockId { + expected: u64, + got: u64, + }, + BrokenChainLink { + expected_prev: HashType, + got_prev: HashType, + }, + HashMismatch { + computed: HashType, + header: HashType, + }, + EmptyBlock, + InvalidClockTransaction, + NonPublicGenesisTransaction, + StateTransition { + /// Index of the failing transaction within the block body. + tx_index: u64, + reason: String, + }, +} + +/// Diagnostic record of the first block that broke the L2 chain. +/// +/// The block-derived fields are `None` for a deserialize break (no header was +/// ever parsed). `l1_slot` is the L1 slot the breaking inscription was read at. +#[derive(Debug, Clone, PartialEq, Eq, Hash, Serialize, Deserialize, JsonSchema)] +pub struct StallReason { + pub block_id: Option, + pub block_hash: Option, + pub prev_block_hash: Option, + pub l1_slot: u64, + pub error: BlockIngestError, + /// The breaking block's L2 timestamp (`None` for a deserialize break). + pub first_seen: Option, + /// Number of later non-chaining blocks seen while the tip is frozen. + pub orphans_since: u64, +} + +/// Status snapshot returned by `getStatus`: the ingestion state plus the +/// indexed L2 tip and, when stalled, the stall diagnostics. +#[derive(Debug, Clone, PartialEq, Eq, Hash, Serialize, Deserialize, JsonSchema)] +pub struct IndexerStatus { + pub state: IndexerSyncState, + pub last_error: Option, + pub indexed_block_id: Option, + pub stall_reason: Option, +} diff --git a/lez/indexer/service/rpc/src/lib.rs b/lez/indexer/service/rpc/src/lib.rs index 5763fe82..8ea807eb 100644 --- a/lez/indexer/service/rpc/src/lib.rs +++ b/lez/indexer/service/rpc/src/lib.rs @@ -1,4 +1,6 @@ -use indexer_service_protocol::{Account, AccountId, Block, BlockId, HashType, Transaction}; +use indexer_service_protocol::{ + Account, AccountId, Block, BlockId, HashType, IndexerStatus, Transaction, +}; use jsonrpsee::proc_macros::rpc; #[cfg(feature = "server")] use jsonrpsee::{core::SubscriptionResult, types::ErrorObjectOwned}; @@ -69,6 +71,9 @@ pub trait Rpc { limit: u64, ) -> Result, ErrorObjectOwned>; + #[method(name = "getStatus")] + async fn get_status(&self) -> Result; + // ToDo: expand healthcheck response into some kind of report #[method(name = "checkHealth")] async fn healthcheck(&self) -> Result<(), ErrorObjectOwned>; diff --git a/lez/indexer/service/src/lib.rs b/lez/indexer/service/src/lib.rs index b1c57163..355d7801 100644 --- a/lez/indexer/service/src/lib.rs +++ b/lez/indexer/service/src/lib.rs @@ -87,6 +87,7 @@ pub async fn run_server( #[cfg(not(feature = "mock-responses"))] let handle = { let service = service::IndexerService::new(config, storage_dir) + .await .context("Failed to initialize indexer service")?; server.start(service.into_rpc()) }; diff --git a/lez/indexer/service/src/mock_service.rs b/lez/indexer/service/src/mock_service.rs index 7bf6c528..70af6239 100644 --- a/lez/indexer/service/src/mock_service.rs +++ b/lez/indexer/service/src/mock_service.rs @@ -10,10 +10,10 @@ use std::{collections::HashMap, sync::Arc, time::Duration}; use indexer_service_protocol::{ Account, AccountId, BedrockStatus, Block, BlockBody, BlockHeader, BlockId, Commitment, - CommitmentSetDigest, Data, EncryptedAccountData, HashType, PrivacyPreservingMessage, - PrivacyPreservingTransaction, ProgramDeploymentMessage, ProgramDeploymentTransaction, - ProgramId, PublicMessage, PublicTransaction, Signature, Transaction, ValidityWindow, - WitnessSet, + CommitmentSetDigest, Data, EncryptedAccountData, HashType, IndexerStatus, IndexerSyncState, + PrivacyPreservingMessage, PrivacyPreservingTransaction, ProgramDeploymentMessage, + ProgramDeploymentTransaction, ProgramId, PublicMessage, PublicTransaction, Signature, + Transaction, ValidityWindow, WitnessSet, }; use jsonrpsee::{ core::{SubscriptionResult, async_trait}, @@ -325,6 +325,24 @@ impl indexer_service_rpc::RpcServer for MockIndexerService { .collect()) } + async fn get_status(&self) -> Result { + let indexed_block_id = self + .state + .read() + .await + .blocks + .iter() + .rev() + .find(|block| block.bedrock_status == BedrockStatus::Finalized) + .map(|block| block.header.block_id); + Ok(IndexerStatus { + state: IndexerSyncState::CaughtUp, + last_error: None, + indexed_block_id, + stall_reason: None, + }) + } + async fn healthcheck(&self) -> Result<(), ErrorObjectOwned> { Ok(()) } diff --git a/lez/indexer/service/src/service.rs b/lez/indexer/service/src/service.rs index 7a8ed90f..767560b1 100644 --- a/lez/indexer/service/src/service.rs +++ b/lez/indexer/service/src/service.rs @@ -4,7 +4,9 @@ use anyhow::{Context as _, Result, bail}; use arc_swap::ArcSwap; use futures::{StreamExt as _, never::Never}; use indexer_core::{IndexerCore, config::IndexerConfig}; -use indexer_service_protocol::{Account, AccountId, Block, BlockId, HashType, Transaction}; +use indexer_service_protocol::{ + Account, AccountId, Block, BlockId, HashType, IndexerStatus, Transaction, +}; use jsonrpsee::{ SubscriptionSink, core::{Serialize, SubscriptionResult, async_trait}, @@ -19,8 +21,8 @@ pub struct IndexerService { } impl IndexerService { - pub fn new(config: IndexerConfig, storage_dir: &Path) -> Result { - let indexer = IndexerCore::new(config, storage_dir)?; + pub async fn new(config: IndexerConfig, storage_dir: &Path) -> Result { + let indexer = IndexerCore::new(config, storage_dir).await?; let subscription_service = SubscriptionService::spawn_new(indexer.clone()); Ok(Self { @@ -149,6 +151,10 @@ impl indexer_service_rpc::RpcServer for IndexerService { Ok(tx_res) } + async fn get_status(&self) -> Result { + Ok(self.indexer.status().into()) + } + async fn healthcheck(&self) -> Result<(), ErrorObjectOwned> { // Checking, that indexer can calculate last state let _ = self diff --git a/lez/sequencer/core/Cargo.toml b/lez/sequencer/core/Cargo.toml index 2931c833..e48f23f2 100644 --- a/lez/sequencer/core/Cargo.toml +++ b/lez/sequencer/core/Cargo.toml @@ -11,6 +11,7 @@ workspace = true lee.workspace = true lee_core.workspace = true common.workspace = true +chain_consistency.workspace = true storage.workspace = true mempool.workspace = true logos-blockchain-zone-sdk.workspace = true diff --git a/lez/sequencer/core/src/block_publisher.rs b/lez/sequencer/core/src/block_publisher.rs index 21551131..b3483336 100644 --- a/lez/sequencer/core/src/block_publisher.rs +++ b/lez/sequencer/core/src/block_publisher.rs @@ -2,14 +2,16 @@ use std::{pin::Pin, sync::Arc, time::Duration}; use anyhow::{Context as _, Result, anyhow}; use common::block::Block; +use futures::StreamExt as _; use log::{info, warn}; pub use logos_blockchain_core::mantle::ops::channel::MsgId; use logos_blockchain_core::mantle::ops::channel::{ChannelId, inscribe::Inscription}; pub use logos_blockchain_key_management_system_service::keys::{Ed25519Key, ZkKey}; pub use logos_blockchain_zone_sdk::sequencer::SequencerCheckpoint; use logos_blockchain_zone_sdk::{ - CommonHttpClient, - adapter::NodeHttpClient, + CommonHttpClient, Slot, ZoneMessage, + adapter::{Node as _, NodeHttpClient}, + indexer::ZoneIndexer, sequencer::{ DepositInfo, Event, FinalizedOp, InscriptionInfo, SequencerConfig as ZoneSdkSequencerConfig, WithdrawArg, WithdrawInfo, ZoneSequencer, @@ -63,12 +65,27 @@ pub trait BlockPublisherTrait: Clone { async fn publish_block(&self, block: &Block, withdrawals: Vec) -> Result<()>; fn channel_id(&self) -> ChannelId; + + /// Current channel frontier slot on the connected chain, or `None` if the + /// channel does not exist there. Drives the startup frontier check. + async fn channel_tip_slot(&self) -> Result>; + + /// Finalized channel messages from `after_slot` (exclusive) up to LIB, used + /// for the startup consistency check and reconstruction. Pass `None` to read + /// from the channel's genesis. + async fn read_channel_after( + &self, + after_slot: Option, + ) -> Result>; } /// Real block publisher backed by zone-sdk's `ZoneSequencer`. #[derive(Clone)] pub struct ZoneSdkPublisher { channel_id: ChannelId, + /// Direct node handle retained for channel reads (startup consistency check + /// and reconstruction); the sequencer itself lives in the drive task. + node: NodeHttpClient, publish_tx: mpsc::Sender<(Inscription, Vec)>, // Aborts the drive task when the last clone is dropped. _drive_task: Arc, @@ -104,7 +121,7 @@ impl BlockPublisherTrait for ZoneSdkPublisher { let mut sequencer = ZoneSequencer::init_with_config( config.channel_id, bedrock_signing_key, - node, + node.clone(), zone_sdk_config, initial_checkpoint, ); @@ -196,6 +213,7 @@ impl BlockPublisherTrait for ZoneSdkPublisher { Ok(Self { channel_id: config.channel_id, + node, publish_tx, _drive_task: Arc::new(DriveTaskGuard(drive_task)), }) @@ -218,6 +236,27 @@ impl BlockPublisherTrait for ZoneSdkPublisher { fn channel_id(&self) -> ChannelId { self.channel_id } + + async fn channel_tip_slot(&self) -> Result> { + Ok(self + .node + .channel_state(self.channel_id) + .await + .context("Failed to read channel state")? + .map(|state| state.tip_slot)) + } + + async fn read_channel_after( + &self, + after_slot: Option, + ) -> Result> { + let indexer = ZoneIndexer::new(self.channel_id, self.node.clone()); + let stream = indexer + .next_messages(after_slot) + .await + .context("Failed to start channel read stream")?; + Ok(stream.collect().await) + } } /// Deserialize inscription payload as a `Block` and return it's`block_id`. diff --git a/lez/sequencer/core/src/block_store.rs b/lez/sequencer/core/src/block_store.rs index aec93d42..4d50c521 100644 --- a/lez/sequencer/core/src/block_store.rs +++ b/lez/sequencer/core/src/block_store.rs @@ -13,7 +13,7 @@ use logos_blockchain_zone_sdk::sequencer::SequencerCheckpoint; pub use storage::DbResult; use storage::sequencer::{ RocksDBIO, - sequencer_cells::{PendingDepositEventRecord, WithdrawalReconciliationKey}, + sequencer_cells::{PendingDepositEventRecord, WithdrawalReconciliationKey, ZoneAnchorRecord}, }; pub struct SequencerStore { @@ -166,6 +166,16 @@ impl SequencerStore { Ok(()) } + /// The last channel block read back and verified from Bedrock (L1 slot + + /// `id`/`hash`), or `None` before any block has been read from the channel. + pub fn get_zone_anchor(&self) -> DbResult> { + self.dbio.get_zone_cursor() + } + + pub fn set_zone_anchor(&self, anchor: &ZoneAnchorRecord) -> DbResult<()> { + self.dbio.put_zone_cursor(anchor) + } + pub fn get_unfulfilled_deposit_events(&self) -> DbResult> { self.dbio.get_pending_deposit_events() } diff --git a/lez/sequencer/core/src/lib.rs b/lez/sequencer/core/src/lib.rs index 75dcc1a9..04a257e5 100644 --- a/lez/sequencer/core/src/lib.rs +++ b/lez/sequencer/core/src/lib.rs @@ -2,17 +2,21 @@ use std::{path::Path, sync::Arc, time::Instant}; use anyhow::{Context as _, Result, anyhow}; use borsh::BorshDeserialize; +use chain_consistency::{Anchor, ChainConsistency, Tip}; use common::{ HashType, block::{BedrockStatus, Block, HashableBlockData}, transaction::{LeeTransaction, clock_invocation}, }; use config::{GenesisAction, SequencerConfig}; -use lee::{AccountId, PublicTransaction, public_transaction::Message}; +use lee::{AccountId, PublicTransaction, V03State, public_transaction::Message}; use lee_core::GENESIS_BLOCK_ID; use log::{error, info, warn}; use logos_blockchain_key_management_system_service::keys::{ED25519_SECRET_KEY_SIZE, Ed25519Key}; -use logos_blockchain_zone_sdk::sequencer::{DepositInfo, WithdrawArg}; +use logos_blockchain_zone_sdk::{ + Slot, ZoneMessage, + sequencer::{DepositInfo, WithdrawArg}, +}; use mempool::{MemPool, MemPoolHandle}; #[cfg(feature = "mock")] pub use mock::SequencerCoreWithMockClients; @@ -20,7 +24,7 @@ use num_bigint::BigUint; pub use storage::error::DbError; use storage::sequencer::{ RocksDBIO, - sequencer_cells::{PendingDepositEventRecord, WithdrawalReconciliationKey}, + sequencer_cells::{PendingDepositEventRecord, WithdrawalReconciliationKey, ZoneAnchorRecord}, }; use crate::{ @@ -64,6 +68,13 @@ pub struct SequencerCore { block_publisher: BP, } +/// Outcome of the startup verify-and-reconstruct pass. +struct ReconstructionSummary { + /// Whether the channel served any blocks. Drives the fresh-start decision + /// between publishing our own genesis and adopting the channel's history. + channel_had_blocks: bool, +} + impl SequencerCore { /// Starts the sequencer using the provided configuration. /// If an existing database is found, the sequencer state is loaded from it and @@ -124,11 +135,7 @@ impl SequencerCore { load_or_create_signing_key(&config.home.join("bedrock_signing_key")) .expect("Failed to load or create bedrock signing key"); - let (store, state, genesis_block) = Self::open_or_create_store(&config); - - let latest_block_meta = store - .latest_block_meta() - .expect("Failed to read latest block meta from store"); + let (mut store, mut state, genesis_block) = Self::open_or_create_store(&config); let initial_checkpoint = store .get_zone_checkpoint() @@ -151,17 +158,27 @@ impl SequencerCore { .await .expect("Failed to initialize Block Publisher"); - // On a truly fresh start (no checkpoint persisted yet), publish the - // genesis block so the indexer can find the channel start. After the - // first publish, zone-sdk's checkpoint persistence covers further - // restarts. - if is_fresh_start { + // Before producing, verify our local state still belongs to the chain + // the channel serves and replay any channel blocks we are missing + // (e.g. from other sequencers). + let reconstruction = Self::verify_and_reconstruct(&block_publisher, &mut store, &mut state) + .await + .expect("Failed to verify/reconstruct sequencer state from Bedrock"); + + // Publish our genesis only when bootstrapping an empty channel. If the + // channel already has blocks (another sequencer bootstrapped it), we + // adopted them during reconstruction instead. + if is_fresh_start && !reconstruction.channel_had_blocks { block_publisher .publish_block(&genesis_block, vec![]) .await .expect("Failed to publish genesis block"); } + let latest_block_meta = store + .latest_block_meta() + .expect("Failed to read latest block meta from store"); + let sequencer_core = Self { state, store, @@ -174,6 +191,155 @@ impl SequencerCore { (sequencer_core, mempool_handle) } + /// Verifies the local store still belongs to the chain the connected channel + /// serves and replays any finalized channel blocks missing locally into + /// `state`/`store`, recording each block's L1 inscription slot as the new + /// anchor. Fails (never parks) on any divergence. + async fn verify_and_reconstruct( + publisher: &BP, + store: &mut SequencerStore, + state: &mut V03State, + ) -> Result { + let anchor_record = store + .get_zone_anchor() + .context("Failed to read zone anchor")?; + + let after_slot = anchor_record + .and_then(|record| record.slot.checked_sub(1)) + .map(Slot::from); + let messages = publisher + .read_channel_after(after_slot) + .await + .context("Failed to read channel history for reconstruction")?; + + // Consistency check against the recorded anchor: fail fast on positive + // evidence the channel is a different chain (missing/behind/reinscribed). + if let Some(record) = anchor_record { + let anchor = Anchor::new( + Slot::from(record.slot), + Some((record.block_id, record.hash)), + ); + let channel_tip_slot = publisher + .channel_tip_slot() + .await + .context("Failed to read channel tip slot")?; + if let ChainConsistency::Inconsistent(mismatch) = + chain_consistency::classify_channel(&anchor, channel_tip_slot, &messages) + { + return Err(anyhow!( + "Sequencer store diverges from the Bedrock channel ({mismatch}). \ + Delete the sequencer storage directory or point at the correct channel." + )); + } + } + + // Replay: apply channel blocks we are missing, verify the ones we have. + let mut channel_had_blocks = false; + for (message, slot) in &messages { + let ZoneMessage::Block(zone_block) = message else { + continue; + }; + let block: Block = borsh::from_slice(&zone_block.data).map_err(|err| { + anyhow!( + "Failed to deserialize channel block at slot {}: {err}", + slot.into_inner() + ) + })?; + channel_had_blocks = true; + Self::apply_reconstructed_block(store, state, &block, *slot)?; + } + + Ok(ReconstructionSummary { channel_had_blocks }) + } + + /// Applies a single channel block during reconstruction: idempotent for + /// blocks we already hold (verifying their hash), a validated continuation + /// for new ones. Advances the persisted anchor to the block's slot. + fn apply_reconstructed_block( + store: &mut SequencerStore, + state: &mut V03State, + block: &Block, + slot: Slot, + ) -> Result<()> { + let tip = store + .latest_block_meta() + .context("Failed to read latest block meta")?; + let block_id = block.header.block_id; + let block_hash = block.header.hash; + + let record = ZoneAnchorRecord { + slot: slot.into_inner(), + block_id, + hash: block_hash, + }; + + // A block at/below the tip must match what we already stored, otherwise + // the channel is a different chain. + if block_id <= tip.id { + match store + .get_block_at_id(block_id) + .context("Failed to read stored block")? + { + Some(stored) if stored.header.hash == block_hash => { + store + .set_zone_anchor(&record) + .context("Failed to persist zone anchor")?; + return Ok(()); + } + Some(stored) => { + return Err(anyhow!( + "Channel block {block_id} hash {block_hash} does not match stored hash {}", + stored.header.hash + )); + } + None => { + return Err(anyhow!( + "Channel block {block_id} is at/below local tip {} but is missing locally", + tip.id + )); + } + } + } + + // New continuation: validate it chains onto the tip, then apply. + let cc_tip = Tip { + block_id: tip.id, + hash: tip.hash, + }; + chain_consistency::validate_against_tip(Some(&cc_tip), block).map_err(|err| { + anyhow!( + "Channel block {block_id} does not extend local tip {}: {err}", + tip.id + ) + })?; + + let mut scratch = state.clone(); + chain_consistency::apply_block(block, &mut scratch) + .map_err(|err| anyhow!("Failed to apply channel block {block_id}: {err}"))?; + + // Derive bridge bookkeeping from the block's transactions, matching the + // production path so the reconciliation counters stay consistent. + let mut deposit_event_ids = Vec::new(); + let mut withdrawals = Vec::new(); + for tx in &block.body.transactions { + if let Some(deposit_id) = extract_bridge_deposit_id(tx) { + deposit_event_ids.push(deposit_id); + } + if let Some(withdraw) = extract_bridge_withdraw_data(tx) { + withdrawals.push(withdraw_event_reconciliation_key(&withdraw.outputs)?); + } + } + + store + .update(block, &deposit_event_ids, withdrawals, &scratch) + .context("Failed to persist reconstructed block")?; + *state = scratch; + store + .set_zone_anchor(&record) + .context("Failed to persist zone anchor")?; + Ok(()) + } + fn on_checkpoint(dbio: Arc) -> block_publisher::CheckpointSink { Box::new(move |cp| { let bytes = match serde_json::to_vec(&cp) { diff --git a/lez/sequencer/core/src/mock.rs b/lez/sequencer/core/src/mock.rs index 39f635f9..65232685 100644 --- a/lez/sequencer/core/src/mock.rs +++ b/lez/sequencer/core/src/mock.rs @@ -4,7 +4,7 @@ use anyhow::Result; use common::block::Block; use logos_blockchain_core::mantle::ops::channel::ChannelId; use logos_blockchain_key_management_system_service::keys::Ed25519Key; -use logos_blockchain_zone_sdk::sequencer::WithdrawArg; +use logos_blockchain_zone_sdk::{Slot, ZoneMessage, sequencer::WithdrawArg}; use crate::{ block_publisher::{ @@ -19,6 +19,28 @@ pub type SequencerCoreWithMockClients = crate::SequencerCore #[derive(Clone)] pub struct MockBlockPublisher { channel_id: ChannelId, + /// Canned channel frontier returned by [`Self::channel_tip_slot`]. + tip_slot: Option, + /// Canned finalized channel history returned by [`Self::read_channel_after`]. + messages: Vec<(ZoneMessage, Slot)>, +} + +impl MockBlockPublisher { + /// Builds a mock publisher backed by a canned channel, for reconstruction + /// and consistency tests. The default (via [`BlockPublisherTrait::new`]) + /// serves an empty channel. + #[must_use] + pub const fn with_canned_channel( + channel_id: ChannelId, + tip_slot: Option, + messages: Vec<(ZoneMessage, Slot)>, + ) -> Self { + Self { + channel_id, + tip_slot, + messages, + } + } } impl BlockPublisherTrait for MockBlockPublisher { @@ -34,6 +56,8 @@ impl BlockPublisherTrait for MockBlockPublisher { ) -> Result { Ok(Self { channel_id: config.channel_id, + tip_slot: None, + messages: Vec::new(), }) } @@ -48,4 +72,21 @@ impl BlockPublisherTrait for MockBlockPublisher { fn channel_id(&self) -> ChannelId { self.channel_id } + + async fn channel_tip_slot(&self) -> Result> { + Ok(self.tip_slot) + } + + async fn read_channel_after( + &self, + after_slot: Option, + ) -> Result> { + // Mirror `next_messages`: `after_slot` is exclusive. + Ok(self + .messages + .iter() + .filter(|(_, slot)| after_slot.is_none_or(|after| *slot > after)) + .cloned() + .collect()) + } } diff --git a/lez/sequencer/core/src/tests.rs b/lez/sequencer/core/src/tests.rs index 4b5ebfbc..9b83344f 100644 --- a/lez/sequencer/core/src/tests.rs +++ b/lez/sequencer/core/src/tests.rs @@ -1232,3 +1232,153 @@ fn pda_mechanism_with_pinata_token_program() { expected_winner_token_holding_post ); } + +mod reconstruction { + use common::block::Block; + use logos_blockchain_core::mantle::ops::channel::{MsgId, inscribe::Inscription}; + use logos_blockchain_zone_sdk::{Slot, ZoneBlock, ZoneMessage}; + use storage::sequencer::sequencer_cells::ZoneAnchorRecord; + + use super::*; + use crate::{SequencerCore, block_store::SequencerStore, mock::MockBlockPublisher}; + + fn block_to_channel_message(block: &Block, slot: u64) -> (ZoneMessage, Slot) { + let bytes = borsh::to_vec(block).expect("serialize block"); + let message = ZoneMessage::Block(ZoneBlock { + id: MsgId::from([0_u8; 32]), + data: Inscription::try_from(bytes.as_slice()).expect("inscription"), + }); + (message, Slot::from(slot)) + } + + /// Collects a sequencer's whole chain (genesis..=tip) into a canned channel, + /// one block per slot at `slot_step` spacing. + fn channel_from_store(store: &SequencerStore, slot_step: u64) -> Vec<(ZoneMessage, Slot)> { + let genesis_id = store.genesis_id(); + let tip_id = store.latest_block_meta().expect("tip").id; + (genesis_id..=tip_id) + .enumerate() + .map(|(index, id)| { + let block = store.get_block_at_id(id).expect("read").expect("present"); + block_to_channel_message(&block, (index as u64 + 1) * slot_step) + }) + .collect() + } + + #[tokio::test] + async fn reconstructs_missing_channel_blocks_into_fresh_store() { + // Sequencer A produces a few blocks; treat its chain as the channel. + let config_a = setup_sequencer_config(); + let (mut seq_a, _handle_a) = + SequencerCoreWithMockClients::start_from_config(config_a.clone()).await; + seq_a.produce_new_block().await.unwrap(); + seq_a.produce_new_block().await.unwrap(); + let tip_a = seq_a.block_store().latest_block_meta().unwrap(); + + let messages = channel_from_store(seq_a.block_store(), 10); + let tip_slot = messages.last().unwrap().1; + let channel_id = config_a.bedrock_config.channel_id; + + // Sequencer B starts from a fresh store and reconstructs A's chain. + let config_b = setup_sequencer_config(); + let (mut store_b, mut state_b, _genesis_b) = + SequencerCore::::open_or_create_store(&config_b); + let mock_b = MockBlockPublisher::with_canned_channel(channel_id, Some(tip_slot), messages); + + let summary = SequencerCore::::verify_and_reconstruct( + &mock_b, + &mut store_b, + &mut state_b, + ) + .await + .expect("reconstruct"); + assert!(summary.channel_had_blocks); + + let tip_b = store_b.latest_block_meta().unwrap(); + assert_eq!(tip_b.id, tip_a.id); + assert_eq!(tip_b.hash, tip_a.hash); + + // State matches: initial account balances agree with sequencer A. + for account in initial_public_user_accounts() { + assert_eq!( + state_b.get_account_by_id(account.account_id).balance, + seq_a.state().get_account_by_id(account.account_id).balance, + ); + } + + let anchor = store_b.get_zone_anchor().unwrap().expect("anchor recorded"); + assert_eq!(anchor.block_id, tip_a.id); + assert_eq!(anchor.slot, tip_slot.into_inner()); + + // Re-running is idempotent: everything is already applied, no error. + let summary = SequencerCore::::verify_and_reconstruct( + &mock_b, + &mut store_b, + &mut state_b, + ) + .await + .expect("reconstruct idempotent"); + assert!(summary.channel_had_blocks); + assert_eq!(store_b.latest_block_meta().unwrap().id, tip_a.id); + } + + #[tokio::test] + async fn fails_when_channel_serves_a_divergent_block() { + let config = setup_sequencer_config(); + let (mut store, mut state, _genesis) = + SequencerCore::::open_or_create_store(&config); + + // Anchor on the local genesis at some slot. + let genesis_id = store.genesis_id(); + let genesis = store.get_block_at_id(genesis_id).unwrap().unwrap(); + let anchor_slot = 100_u64; + store + .set_zone_anchor(&ZoneAnchorRecord { + slot: anchor_slot, + block_id: genesis_id, + hash: genesis.header.hash, + }) + .unwrap(); + + // The channel serves a different block at the anchor id/slot. + let mut tampered = genesis.clone(); + tampered.header.hash = HashType([9_u8; 32]); + let messages = vec![block_to_channel_message(&tampered, anchor_slot)]; + let mock = MockBlockPublisher::with_canned_channel( + config.bedrock_config.channel_id, + Some(Slot::from(anchor_slot)), + messages, + ); + + let result = SequencerCore::::verify_and_reconstruct( + &mock, &mut store, &mut state, + ) + .await; + assert!(result.is_err(), "divergent channel must abort startup"); + } + + #[tokio::test] + async fn fails_when_channel_is_missing() { + let config = setup_sequencer_config(); + let (mut store, mut state, _genesis) = + SequencerCore::::open_or_create_store(&config); + let genesis_id = store.genesis_id(); + let genesis = store.get_block_at_id(genesis_id).unwrap().unwrap(); + store + .set_zone_anchor(&ZoneAnchorRecord { + slot: 100, + block_id: genesis_id, + hash: genesis.header.hash, + }) + .unwrap(); + + // Anchor present, but the channel does not exist on the connected chain. + let mock = + MockBlockPublisher::with_canned_channel(config.bedrock_config.channel_id, None, vec![]); + let result = SequencerCore::::verify_and_reconstruct( + &mock, &mut store, &mut state, + ) + .await; + assert!(result.is_err(), "missing channel must abort startup"); + } +} diff --git a/lez/storage/src/indexer/indexer_cells.rs b/lez/storage/src/indexer/indexer_cells.rs index b19a5510..d71a2ed6 100644 --- a/lez/storage/src/indexer/indexer_cells.rs +++ b/lez/storage/src/indexer/indexer_cells.rs @@ -8,8 +8,8 @@ use crate::{ indexer::{ ACC_NUM_CELL_NAME, BLOCK_HASH_CELL_NAME, BREAKPOINT_CELL_NAME, CF_ACC_META, CF_BREAKPOINT_NAME, CF_HASH_TO_ID, CF_TX_TO_ID, DB_META_LAST_BREAKPOINT_ID, - DB_META_LAST_OBSERVED_L1_LIB_HEADER_ID_IN_DB_KEY, DB_META_ZONE_SDK_INDEXER_CURSOR_KEY, - TX_HASH_CELL_NAME, + DB_META_LAST_OBSERVED_L1_LIB_HEADER_ID_IN_DB_KEY, DB_META_STALL_REASON_KEY, + DB_META_ZONE_SDK_INDEXER_CURSOR_KEY, TX_HASH_CELL_NAME, }, }; @@ -247,6 +247,40 @@ impl SimpleWritableCell for ZoneSdkIndexerCursorCellRef<'_> { } } +/// Opaque JSON bytes for the indexer's persisted `Option`. +#[derive(BorshDeserialize)] +pub struct StallReasonCellOwned(pub Vec); + +impl SimpleStorableCell for StallReasonCellOwned { + type KeyParams = (); + + const CELL_NAME: &'static str = DB_META_STALL_REASON_KEY; + const CF_NAME: &'static str = CF_META_NAME; +} + +impl SimpleReadableCell for StallReasonCellOwned {} + +#[derive(BorshSerialize)] +pub struct StallReasonCellRef<'bytes>(pub &'bytes [u8]); + +impl SimpleStorableCell for StallReasonCellRef<'_> { + type KeyParams = (); + + const CELL_NAME: &'static str = DB_META_STALL_REASON_KEY; + const CF_NAME: &'static str = CF_META_NAME; +} + +impl SimpleWritableCell for StallReasonCellRef<'_> { + fn value_constructor(&self) -> DbResult> { + borsh::to_vec(&self).map_err(|err| { + DbError::borsh_cast_message( + err, + Some("Failed to serialize stall reason cell".to_owned()), + ) + }) + } +} + #[cfg(test)] mod uniform_tests { use crate::{ diff --git a/lez/storage/src/indexer/mod.rs b/lez/storage/src/indexer/mod.rs index 87d586fb..10c80e63 100644 --- a/lez/storage/src/indexer/mod.rs +++ b/lez/storage/src/indexer/mod.rs @@ -24,6 +24,8 @@ pub const DB_META_LAST_OBSERVED_L1_LIB_HEADER_ID_IN_DB_KEY: &str = pub const DB_META_LAST_BREAKPOINT_ID: &str = "last_breakpoint_id"; /// Key base for storing the zone-sdk indexer cursor (opaque bytes). pub const DB_META_ZONE_SDK_INDEXER_CURSOR_KEY: &str = "zone_sdk_indexer_cursor"; +/// Key base for storing the persisted `Option` diagnostic record (opaque JSON bytes). +pub const DB_META_STALL_REASON_KEY: &str = "stall_reason"; /// Cell name for a breakpoint. pub const BREAKPOINT_CELL_NAME: &str = "breakpoint"; @@ -211,20 +213,7 @@ fn apply_block_transactions(mut block: Block, state: &mut V03State) -> DbResult< })?; } else { transaction - .transaction_stateless_check() - .map_err(|err| { - DbError::db_interaction_error(format!( - "transaction pre check failed with err {err:?}" - )) - })? - // FIXME: HOT FIX (testnet v0.2): does not check for system account updates due to - // sequencer-generated deposit tx'es; - // CHANGE ME back to `execute_check_on_state` when the indexer can authenticate deposit transactions - .execute_without_system_accounts_check_on_state( - state, - block.header.block_id, - block.header.timestamp, - ) + .execute_on_state(state, block.header.block_id, block.header.timestamp) .map_err(|err| { DbError::db_interaction_error(format!( "transaction execution failed with err {err:?}" diff --git a/lez/storage/src/indexer/read_once.rs b/lez/storage/src/indexer/read_once.rs index 6e79adc4..591f8405 100644 --- a/lez/storage/src/indexer/read_once.rs +++ b/lez/storage/src/indexer/read_once.rs @@ -4,7 +4,8 @@ use crate::{ cells::shared_cells::{BlockCell, FirstBlockCell, FirstBlockSetCell, LastBlockCell}, indexer::indexer_cells::{ AccNumTxCell, BlockHashToBlockIdMapCell, BreakpointCellOwned, LastBreakpointIdCell, - LastObservedL1LibHeaderCell, TxHashToBlockIdMapCell, ZoneSdkIndexerCursorCellOwned, + LastObservedL1LibHeaderCell, StallReasonCellOwned, TxHashToBlockIdMapCell, + ZoneSdkIndexerCursorCellOwned, }, }; @@ -73,4 +74,8 @@ impl RocksDBIO { .get_opt::(())? .map(|cell| cell.0)) } + + pub fn get_stall_reason_bytes(&self) -> DbResult>> { + Ok(self.get_opt::(())?.map(|cell| cell.0)) + } } diff --git a/lez/storage/src/indexer/write_non_atomic.rs b/lez/storage/src/indexer/write_non_atomic.rs index 7ddab1dd..215250b7 100644 --- a/lez/storage/src/indexer/write_non_atomic.rs +++ b/lez/storage/src/indexer/write_non_atomic.rs @@ -3,7 +3,7 @@ use crate::{ DBIO as _, cells::shared_cells::{FirstBlockSetCell, LastBlockCell}, indexer::indexer_cells::{ - BreakpointCellRef, LastBreakpointIdCell, LastObservedL1LibHeaderCell, + BreakpointCellRef, LastBreakpointIdCell, LastObservedL1LibHeaderCell, StallReasonCellRef, ZoneSdkIndexerCursorCellRef, }, }; @@ -35,6 +35,10 @@ impl RocksDBIO { self.put(&ZoneSdkIndexerCursorCellRef(bytes), ()) } + pub fn put_stall_reason_bytes(&self, bytes: &[u8]) -> DbResult<()> { + self.put(&StallReasonCellRef(bytes), ()) + } + // State pub fn put_breakpoint(&self, br_id: u64, breakpoint: &V03State) -> DbResult<()> { diff --git a/lez/storage/src/sequencer/mod.rs b/lez/storage/src/sequencer/mod.rs index 44068517..13fcebdd 100644 --- a/lez/storage/src/sequencer/mod.rs +++ b/lez/storage/src/sequencer/mod.rs @@ -20,7 +20,7 @@ use crate::{ LEEStateCellOwned, LEEStateCellRef, LastFinalizedBlockIdCell, LatestBlockMetaCellOwned, LatestBlockMetaCellRef, PendingDepositEventRecord, PendingDepositEventsCellOwned, PendingDepositEventsCellRef, UnseenWithdrawCountCell, WithdrawalReconciliationKey, - ZoneSdkCheckpointCellOwned, ZoneSdkCheckpointCellRef, + ZoneAnchorRecord, ZoneCursorCell, ZoneSdkCheckpointCellOwned, ZoneSdkCheckpointCellRef, }, }; @@ -32,6 +32,10 @@ pub const DB_META_LAST_FINALIZED_BLOCK_ID: &str = "last_finalized_block_id"; pub const DB_META_LATEST_BLOCK_META_KEY: &str = "latest_block_meta"; /// Key base for storing the zone-sdk sequencer checkpoint (opaque bytes). pub const DB_META_ZONE_SDK_CHECKPOINT_KEY: &str = "zone_sdk_checkpoint"; +/// Key base for storing the last channel block read back and verified from +/// Bedrock (its L1 slot + `id`/`hash`) — the anchor for the startup +/// consistency check and the resume point for reconstruction. +pub const DB_META_ZONE_CURSOR_KEY: &str = "zone_cursor"; /// Key base for storing queued deposit events that were not yet /// fulfilled on L2. pub const DB_META_PENDING_DEPOSIT_EVENTS_KEY: &str = "pending_deposit_events"; @@ -246,6 +250,14 @@ impl RocksDBIO { self.put(&ZoneSdkCheckpointCellRef(bytes), ()) } + pub fn get_zone_cursor(&self) -> DbResult> { + Ok(self.get_opt::(())?.map(|cell| cell.0)) + } + + pub fn put_zone_cursor(&self, anchor: &ZoneAnchorRecord) -> DbResult<()> { + self.put(&ZoneCursorCell(*anchor), ()) + } + pub fn get_pending_deposit_events(&self) -> DbResult> { Ok(self .get_opt::(())? diff --git a/lez/storage/src/sequencer/sequencer_cells.rs b/lez/storage/src/sequencer/sequencer_cells.rs index 7672e271..302a0b97 100644 --- a/lez/storage/src/sequencer/sequencer_cells.rs +++ b/lez/storage/src/sequencer/sequencer_cells.rs @@ -9,7 +9,8 @@ use crate::{ sequencer::{ CF_LEE_STATE_NAME, DB_LEE_STATE_KEY, DB_META_LAST_FINALIZED_BLOCK_ID, DB_META_LATEST_BLOCK_META_KEY, DB_META_PENDING_DEPOSIT_EVENTS_KEY, - DB_META_UNSEEN_WITHDRAW_COUNT_KEY, DB_META_ZONE_SDK_CHECKPOINT_KEY, + DB_META_UNSEEN_WITHDRAW_COUNT_KEY, DB_META_ZONE_CURSOR_KEY, + DB_META_ZONE_SDK_CHECKPOINT_KEY, }, }; @@ -132,6 +133,39 @@ impl SimpleWritableCell for ZoneSdkCheckpointCellRef<'_> { } } +/// The last channel block read back and verified from Bedrock. +/// +/// Holds its L1 inscription `slot` plus the block's `id`/`hash`, and serves as +/// both the anchor for the startup consistency check and the resume point for +/// reconstruction. `slot` is stored as a raw `u64` because the zone-sdk `Slot` +/// does not derive borsh; the caller converts to/from `Slot`. +#[derive(Debug, Clone, Copy, BorshSerialize, BorshDeserialize)] +pub struct ZoneAnchorRecord { + pub slot: u64, + pub block_id: u64, + pub hash: HashType, +} + +#[derive(Debug, BorshSerialize, BorshDeserialize)] +pub struct ZoneCursorCell(pub ZoneAnchorRecord); + +impl SimpleStorableCell for ZoneCursorCell { + type KeyParams = (); + + const CELL_NAME: &'static str = DB_META_ZONE_CURSOR_KEY; + const CF_NAME: &'static str = CF_META_NAME; +} + +impl SimpleReadableCell for ZoneCursorCell {} + +impl SimpleWritableCell for ZoneCursorCell { + fn value_constructor(&self) -> DbResult> { + borsh::to_vec(&self).map_err(|err| { + DbError::borsh_cast_message(err, Some("Failed to serialize zone cursor".to_owned())) + }) + } +} + #[derive(Debug, Clone, PartialEq, Eq, BorshSerialize, BorshDeserialize)] pub struct PendingDepositEventRecord { pub deposit_op_id: HashType, diff --git a/test_fixtures/src/config.rs b/test_fixtures/src/config.rs index 92ed6a18..4755a2f1 100644 --- a/test_fixtures/src/config.rs +++ b/test_fixtures/src/config.rs @@ -176,6 +176,7 @@ pub fn indexer_config(bedrock_addr: SocketAddr) -> Result { auth: None, }, channel_id: bedrock_channel_id(), + allow_chain_reset: false, }) }