Skip to main content

Operations

Operational guidance for running Excalibur in production environments, including resilience, recovery procedures, and maintenance runbooks.

Before You Start​

Guides​

TopicDescription
Runtime ContractCanonical runtime semantics for dispatch ordering, cancellation, retries, and context propagation
Reliability GuaranteesDelivery/ordering/deduplication/dead-letter guarantees by execution path and provider family
SLO, SLI, and TelemetryProduction objectives and telemetry schema for release readiness and operations
Incident RunbooksEscalation model and step-by-step response playbooks for common runtime incidents
Operational ResilienceTransient error handling, retry policies, and recovery strategies
Recovery RunbooksStep-by-step recovery procedures for common failure scenarios

Quick Reference​

Provider Resilience Matrix​

ProviderRetry PolicyRecovery OptionsCDC Position Recovery
SQL ServerSqlServerRetryPolicyAutomatic reconnectCdcRecoveryOptions
PostgreSQLPostgresRetryPolicyAutomatic reconnectPostgresCdcRecoveryOptions
CosmosDBSDK-managedAutomaticContinuation token
DynamoDBSDK-managedAutomaticStream ARN
MongoDBDriver poolAutomaticResume token
RedisManual reconnectConnectionMultiplexerN/A

Key Error Codes​

SQL Server Transient Errors:

  • 596 - Session killed by backup/restore (critical for CDC)
  • 9001, 9002 - Transaction log unavailable
  • 1205 - Deadlock victim
  • 40613 - Database unavailable

PostgreSQL Transient Errors:

  • 08xxx - Connection errors
  • 40001, 40P01 - Serialization/deadlock
  • 57Pxx - Admin/crash shutdown
  • 53xxx - Insufficient resources

See Also​