Implement#

Write-Ahead Log and Event Bus Connector

Overview#

Implement durable event persistence and broker connectivity for
procmond. This ticket establishes the foundation for reliable event
delivery with crash recovery by creating the Write-Ahead Log (WAL)
component and EventBusConnector that integrates with
daemoneye-eventbus.

Scope#

In Scope:

• WriteAheadLog component with bincode serialization
• Sequence-numbered WAL files with rotation at 80% capacity
• CRC32 corruption detection and recovery
• EventBusConnector with WAL integration
• Event buffering (10MB limit) with replay capability
• Connection to broker via DAEMONEYE_BROKER_SOCKET
  environment variable
• Dynamic backpressure monitoring (70% threshold)
• Unit tests for WAL and EventBusConnector
  Out of Scope:
• Actor pattern implementation (Ticket 2)
• RPC service handling (Ticket 3)
• Agent-side changes (Ticket 4)
• Integration testing (Ticket 5)

Technical Details#

WriteAheadLog Component#

Location: file:procmond/src/wal.rs
Key Responsibilities:

• Persist events to disk before buffering (durability guarantee)
• Use sequence-numbered files:
  procmond-{sequence:05}.wal
• Rotate when file reaches 80MB (80% of 100MB max)
• Replay events on startup for crash recovery
• Delete WAL files after successful publish
• Handle corruption with CRC32 validation
  File Format:
  // Bincode-serialized records with CRC32 checksumsstruct WalEntry { sequence:u64, timestamp: DateTime<Utc>, event: ProcessEvent, crc32:u32,}

EventBusConnector Component#

Location:
file:procmond/src/event_bus_connector.rs
Key Responsibilities:

• Connect to daemoneye-agent's embedded broker

• Integrate with WriteAheadLog for event persistence

• Buffer events (10MB limit) when connection lost

• Replay buffered events from WAL on reconnection

• Publish to topic hierarchy: events.process.*

• Monitor buffer level for backpressure (70% threshold)

• Provide shared channel reference for backpressure signaling
  Backpressure Strategy:

• Trigger at 70% buffer capacity

• Release at 50% buffer capacity

• Signal via shared mpsc channel (to be used by Ticket 2)
  `sequenceDiagram
  participant WAL as WriteAheadLog
  participant Connector as EventBusConnector
  participant Broker as DaemoneyeBroker

  Note over WAL,Broker: Normal Operation
  Connector->>WAL: Write event to WAL
  WAL-->>Connector: Persisted (sequence number)
  Connector->>Broker: Publish event
  Broker-->>Connector: Acknowledged
  Connector->>WAL: Mark for deletion

  Note over WAL,Broker: Connection Lost
  Connector->>WAL: Write event to WAL
  WAL-->>Connector: Persisted
  Connector->>Connector: Buffer event (in-memory)
  Note over Connector: Buffer reaches 70%
  Connector->>Connector: Trigger backpressure

  Note over WAL,Broker: Reconnection
  Connector->>Broker: Reconnect
  Connector->>WAL: Read unpublished events
  WAL-->>Connector: Event list
  Connector->>Broker: Replay events
  Broker-->>Connector: Acknowledged
  Connector->>WAL: Delete WAL files
  Note over Connector: Buffer drops below 50%
  Connector->>Connector: Release backpressure`

Dependencies#

Requires:

• daemoneye-eventbus client library
• Existing ProcessEvent data model from
  file:daemoneye-lib/src/models/process.rs
  Blocks:
• ticket:54226c8a-719a-479a-863b-9c91f43717a9/[Ticket 2] - Actor
  pattern needs EventBusConnector
• ticket:54226c8a-719a-479a-863b-9c91f43717a9/[Ticket 3] - RPC service
  needs event bus connectivity

Acceptance Criteria#

WriteAheadLog#

• Events persisted to disk using
  bincode serialization
• Sequence-numbered files created:
  procmond-00001.wal, procmond-00002.wal,
  etc.
• File rotation occurs at 80MB (80% of
  100MB max)
• WAL replay works correctly on
  startup (all unpublished events recovered)
• Corrupted entries detected via CRC32
  and skipped with warning log
• WAL files deleted after all events
  successfully published
• Unit tests cover: persistence,
  rotation, replay, corruption recovery

EventBusConnector#

• Connects to broker using
  DAEMONEYE_BROKER_SOCKET environment variable
• Events written to WAL before
  buffering
• Events buffered (10MB limit) when
  connection lost
• Buffered events replayed on
  reconnection
• Events published to correct topics:
  events.process.start, events.process.stop,
  events.process.modify
• Backpressure triggered at 70% buffer
  capacity
• Backpressure released at 50% buffer
  capacity
• Shared channel reference provided
  for backpressure signaling
• Unit tests cover: connection, WAL
  integration, buffering, replay, backpressure

Integration#

• EventBusConnector successfully
  integrates with WriteAheadLog
• Events survive procmond crash and
  are replayed on restart
• No data loss during connection
  failures
• Backpressure mechanism ready for
  actor integration (Ticket 2)

References#

• Epic Brief:
  spec:54226c8a-719a-479a-863b-9c91f43717a9/0fc3298b-37df-4722-a761-66a5a0da16b3
• Tech Plan:
  spec:54226c8a-719a-479a-863b-9c91f43717a9/f70103e2-e7ef-494f-8638-5a7324565f28
  (Phase 1, Component Architecture)
• Event Bus Architecture:
  file/embedded-broker-architecture.md
• Topic Hierarchy:
  file/docs/topic-hierarchy.md
• Process Models:
  file/src/models/process.rs

Source note: Migrated from the public repo
(spec/procmond/tickets/Implement_Write-Ahead_Log_and_Event_Bus_Connector.md)
on 2026-04-17. The repo copy has been removed.