Topics properties
Generated from the connector's ConfigDef by ConfigDocsGeneratorTest; do not edit by hand.
| Property | Type | Default | Importance | Description |
|---|---|---|---|---|
cdc.topic.prefix | string | required | high | Prefix of every topic this connector writes; also the logical server name in the Debezium-compatible envelope and the offset partition. |
cdc.topic.template | string | none | medium | Topic name template. Variables: ${prefix}, ${pdb}, ${schema}, ${table}, ${database}. Default ${prefix}.${pdb}.${schema}.${table} in a CDB and ${prefix}.${schema}.${table} otherwise. Characters Kafka does not allow become underscores. Tables the template names differently that end up on one topic only through that replacement (APP.ORDER# and APP.ORDER$ both become ORDER_) stop the task with CDC-6004 before either is published; a template that leaves out ${table} or ${schema} on purpose is not affected. |
cdc.tables.include | list | required | high | Comma-separated regular expressions over PDB.SCHEMA.TABLE (CDB) or SCHEMA.TABLE (non-CDB) selecting the tables to capture. |
cdc.tables.exclude | list | empty | medium | Comma-separated regular expressions removing tables from the included set. |
cdc.tables.case.sensitive | boolean | false | low | Match table and column patterns case-sensitively. Oracle stores unquoted names in upper case. |
cdc.columns.exclude | list | empty | medium | Comma-separated regular expressions over PDB.SCHEMA.TABLE.COLUMN (CDB) or SCHEMA.TABLE.COLUMN (non-CDB) naming columns to leave out of every record, matched against the whole name and case-insensitively unless cdc.tables.case.sensitive=true. An excluded column is dropped while its row is decoded, before its value is converted, so it never reaches a record, a Connect schema, the transaction buffer, the spill files, the transaction journal or a log line; snapshots do not select it and reselect never fetches it. Rows of a table these patterns may match go to the DLQ without their SQL_REDO and SQL_UNDO, and a statement that fails to parse is reported without quoting it. A column of the record key (primary key, unique index or cdc.key.columns) cannot be excluded: validation reports it, and a task that finds one at start or after a DDL stops with CDC-3001. A column added later that matches is excluded from its first row. |
cdc.users.exclude | list | empty | low | Oracle users whose transactions are dropped in the mining query, so they never create transactions in the buffer (for example a replication or GoldenGate user). |
cdc.key.missing | string | fail | medium | What to do with a captured table that has neither a primary key nor a NOT NULL unique index: fail (at validation), rowid (key records by ROWID; a moved row changes its key) or none (no key). |
cdc.key.columns | string | empty | low | Per-table key override as SCHEMA.TABLE:COL1,COL2;SCHEMA.OTHER:COL, taking precedence over the primary key. |
cdc.tombstones.on.delete | boolean | true | medium | Emit a null-valued record after every delete so compacted topics drop the key. |
cdc.ops.topic | string | ${prefix}.cdc.ops | low | Topic for the connector's operational events (task start and stop, DDL seen, decode failures, reconnects, discarded or released transactions, signal acknowledgements); ${prefix} expands to the topic prefix. |
cdc.signals.topic | string | ${prefix}.cdc.signals | low | Topic the connector reads signals from (snapshot, refresh-tables, log-state). |
cdc.schema.topic | string | ${prefix}.cdc.schema | low | Compacted topic holding table schema versions. |
cdc.dlq.topic | string | ${prefix}.cdc.dlq | low | Dead letter topic for rows that could not be decoded or that LogMiner marked unsupported (used only with cdc.on.decode.error=dlq) and for transactions discarded by cdc.transaction.max.age.action=discard. Records carry the raw SQL_REDO, SCN, XID, table and exception; for a table whose columns cdc.columns.exclude may match, the SQL_REDO and SQL_UNDO are withheld. |
cdc.transactions.topic.enabled | boolean | false | low | Creates the transaction metadata topic when the connector has broker access. Writing BEGIN and END records per Oracle transaction to it is not built yet, so the topic stays empty. |
cdc.transactions.topic | string | ${prefix}.cdc.transactions | low | Transaction metadata topic (see cdc.transactions.topic.enabled). |
cdc.kafka.bootstrap.servers | string | none | medium | Bootstrap servers for the connector's own Kafka clients: the admin client that creates the internal topics with the right cleanup policy, and the readers of the schema and journal topics. Other client settings go under cdc.kafka.*. When unset the connector relies on the worker's topic creation and on pre-created topics. |
cdc.internal.topic.replication.factor | short | -1 | low | Replication factor for internal topics the connector creates; -1 uses the broker default. |
cdc.journal.converter | string | sh.oso.connect.oracle.journal.TolerantJsonConverter | low | Converter the task uses to read the transaction journal and schema topics back at start. The default reads JSON written with or without the schema envelope, so it matches a worker using the JSON converter in either mode. Set it to the worker's converter class when the worker uses another converter (Avro, Protobuf); settings for it go under cdc.journal.converter.*. |
cdc.heartbeat.interval.ms | long | 10000 | medium | How often a heartbeat record carrying the current position is written when no change records flow, so Connect commits offsets on a quiet database and the start position of a new connector becomes durable at once. 0 disables periodic heartbeats; the start heartbeat is always written. |
cdc.heartbeat.topic | string | ${prefix}.cdc.heartbeat | low | Topic for heartbeat records; ${prefix} expands to the topic prefix. |