Skip to content

Latest commit

 

History

History
160 lines (135 loc) · 8.14 KB

configuration.md

File metadata and controls

160 lines (135 loc) · 8.14 KB

Configuration Reference

← Back to docs

Configuration is loaded from node.yaml (node-level settings). Per-table configuration and assignments are stored in the database.

node.yaml — Node Settings

Settings are organized by role: shared resources (database, storage), network (grpc, health), coordinator logic (coordinator), and worker pool (worker). Per-table configuration (Kafka routing, feature flags) is managed via the admin gRPC API and stored in the database.

database:
  # Primary (RW). Required for coordinator and worker roles.
  primary:
    host: localhost
    port: 3306                     # Required
    database: metalog_metastore
    user: root
    password: ""
    poolSize: 5                    # Max open connections (default: 5)
    poolMinIdle: 2                 # Min idle connections (default: 2)

  # Replica (RO, optional). Query and metadata services use this pool when
  # configured. Falls back to primary if omitted. For replica-only deployments
  # (API server), omit primary entirely and configure only replica.
  # replica:
  #   host: replica-db
  #   port: 3306
  #   database: metalog_metastore
  #   user: reader
  #   password: secret
  #   poolSize: 10
  #   poolMinIdle: 2

storage:
  defaultBackend: minio
  backends:
    minio:
      endpoint: http://localhost:9000
      accessKey: minioadmin
      secretKey: minioadmin
      region: ""                 # AWS region (optional, for S3)
      bucket: logs               # Storage bucket for this backend
      forcePathStyle: true       # Required for MinIO

health:
  enabled: true
  port: 8081                     # Default: 8081

grpc:
  port: 9090                     # Default: 9090
  ingestion: true                # Requires database.primary
  admin: true                    # Requires database.primary
  query: true                    # Uses database.replica, falls back to primary
  metadata: true                 # Uses database.replica, falls back to primary

coordinator:
  enabled: true                      # Enable coordinator subsystem
  nodeIdEnvVar: HOSTNAME         # Env var whose value becomes node_id in _table_assignment
                                 # Falls back to os.Hostname()
  reconciliationIntervalSeconds: 60  # Default: 60. How often to reconcile table assignments.

  # HA settings
  haStrategy: heartbeat              # "heartbeat" (default) or "lease"
  heartbeatIntervalSeconds: 30       # Default: 30. Heartbeat mode: liveness write interval.
  deadNodeThresholdSeconds: 180      # Default: 180. Heartbeat mode: seconds before node is declared dead.
  leaseTtlSeconds: 180               # Default: 180. Lease mode: lease duration.
  leaseRenewalIntervalSeconds: 30    # Default: 30. Lease mode: renewal interval (must be < leaseTtlSeconds).

# Tables are registered via admin gRPC API (AdminService/RegisterTable).
# The coordinator discovers assigned tables from the DB on startup and via
# periodic reconciliation. See docs/guides/configure-tables.md.

# Shared worker pool (claims tasks from all tables).
worker:
  concurrency: 4                 # 0 = workers disabled
  # clpBinaryPath: /usr/bin/clp-s  # Auto-resolved from $PATH if omitted
  # clpProcessTimeoutSeconds: 300  # Default: 300 (5 minutes)

Configuration Reference Table

Property Default Description
database.primary.host — Primary (RW) database hostname. Required for coordinator/worker roles
database.primary.port — Required. Primary port (1-65535)
database.primary.database — Primary database name
database.primary.user — Primary database user
database.primary.password — Primary database password
database.primary.poolSize 5 Primary max open connections
database.primary.poolMinIdle 2 Primary min idle connections
database.replica.host — Replica (RO) database hostname. Optional; falls back to primary
database.replica.port — Required. Replica port (1-65535)
database.replica.database — Replica database name
database.replica.user — Replica database user
database.replica.password — Replica database password
database.replica.poolSize 5 Replica max open connections
database.replica.poolMinIdle 2 Replica min idle connections
storage.defaultBackend — Default storage backend name
health.enabled false Enable HTTP health endpoint
health.port 8081 Health endpoint port
grpc.port 9090 gRPC server port (server starts if any service is enabled)
grpc.ingestion false Enable ingestion gRPC service (requires database.primary)
grpc.admin false Enable admin gRPC service (requires database.primary)
grpc.query false Enable query gRPC service (uses database.replica, falls back to primary)
grpc.metadata false Enable metadata gRPC service (uses database.replica, falls back to primary)
coordinator.enabled false Enable coordinator subsystem
coordinator.nodeIdEnvVar HOSTNAME Env var for node identity
coordinator.reconciliationIntervalSeconds 60 Table assignment reconciliation interval
coordinator.haStrategy heartbeat HA mode: heartbeat or lease
coordinator.heartbeatIntervalSeconds 30 Heartbeat mode: liveness write interval
coordinator.deadNodeThresholdSeconds 180 Heartbeat mode: seconds before node is dead
coordinator.leaseTtlSeconds 180 Lease mode: lease duration in seconds
coordinator.leaseRenewalIntervalSeconds 30 Lease mode: renewal interval (must be < TTL)
worker.concurrency 0 (disabled) Concurrent task goroutines
worker.clpBinaryPath $PATH lookup Path to clp-s binary. Auto-resolved from $PATH if omitted
worker.clpProcessTimeoutSeconds 300 CLP process timeout

Internal Constants

These are not configurable via YAML but are defined in internal/config/timeouts.go:

Constant Value Description
DefaultBatchFlushInterval 1s Batching writer flush interval
DefaultBatchSize 5000 UPSERT batch size
DefaultTaskClaimBatchSize 10 Tasks claimed per worker poll
DefaultTaskStaleTimeout 5m Processing task reclaim timeout
DefaultTaskCleanupAge 24h Completed/failed task cleanup age
DefaultWorkerPollInterval 2s Worker task poll interval
DefaultWorkerBackoffMax 30s Max backoff when no tasks available
DefaultPlannerInterval 60s Consolidation planner run interval
defaultScanInterval 60s Retention strategy scan interval
defaultDeleteRate 500/sec Max storage object deletions per second during retention cleanup
DefaultQueryLimit 1000 Default total result limit for split queries (0 = unlimited)
DefaultQueryPageSize 1000 Internal SQL LIMIT per page for streaming queries
DefaultDeadlockMaxRetries 10 Max retries on SQL deadlock

Schema Auto-Creation

All schema setup is automatic. On startup, the node creates the database, initializes the _table* registry and _task_queue, and creates per-table metadata tables. See Architecture Overview for the full startup sequence.

See Also