CDC Sink: Server Configuration

• This article documents the RavenDB server configuration keys that control CDC Sink task behavior.

• These keys can be set in settings.json or passed as environment variables.
  All CDC Sink keys have a Server-wide or per database scope, so they can be set per database as well as server-wide. Learn how to apply these keys in the Configuration overview article.

• In this article:

  • CdcSink.MaxBatchSize
  • CdcSink.MaxFallbackTimeInSec
  • CdcSink.SqlServer.PollIntervalInSec
  • CdcSink.Postgres.ReplicationTimeoutInSec

CdcSink.MaxBatchSize

Target number of change rows processed in a single batch before writing to RavenDB.
A batch may exceed this size when a source database transaction contains more rows,
since transactions are never split across batches.

Larger values increase throughput but also increase memory usage per batch.

• Type: int
• Default: 1024
• Scope: Server-wide or per database

CdcSink.MaxFallbackTimeInSec

Maximum number of seconds used as the base cap for CDC Sink fallback mode after a failure.
The first retry after a failure waits 5 seconds.
Subsequent retries roughly double based on the time since the previous error, up to this configured base cap.

A small random jitter is added after the base delay to avoid synchronized retries across multiple tasks,
so the actual wait can be slightly higher than this value.

• Type: int
• Default: 900 (15 minutes)
• Scope: Server-wide or per database

CdcSink.SqlServer.PollIntervalInSec

How frequently (in seconds) the SQL Server CDC Sink polls for new change rows.
Lower values reduce latency but increase load on the source database.

PostgreSQL and MySQL use streaming replication and ignore this setting.

• Type: int
• Default: 1
• Scope: Server-wide or per database

CdcSink.Postgres.ReplicationTimeoutInSec

Timeout (in seconds) for the PostgreSQL replication connection.
Controls both the server-side wal_sender_timeout (keepalives arrive at roughly half this interval) and the client-side WalReceiverTimeout.

Lower values detect dead connections faster but increase keepalive traffic.

SQL Server and MySQL ignore this setting.

• Type: int
• Default: 10
• Scope: Server-wide or per database