Debezium lost one message after OLR restart

[JJaneq]

Bug reporter information

Full name
Jan Bejgier

Email or GitHub username
jan.bejgier@goldenore.com

Company or organization
Goldenore Sp. z o.o.

---

Bug description

When OLR is restarted during data transmission to the Debezium connector, one message is lost.

OLR, after receiving a CONTINUE request with a given c_scn and c_idx, resumes streaming data but always skips message that is immediately following the requested c_idx.

---

Step-by-step reproduction instructions

1. Clone the required repository:

   git clone https://github.com/bersler/OpenLogReplicator-tutorials.git

2. Create the required images:

   cd OpenLogReplicator-tutorials/images/
   ./oracle_database_21.3.0-xe.sh
   ./bersler_openlogreplicator_tutorial_dev.sh

3. Start the Oracle database with modified compose.yaml:

   services:
      oracle:
         image: "${DB_IMAGE}"
         container_name: ORA1
         volumes:
            - ./fra:/opt/oracle/fra
            - ./oradata:/opt/oracle/oradata
            - ./sql:/opt/sql
            - ./setup:/opt/oracle/scripts/setup
         networks:
            - isolated_network
         environment:
            - ORACLE_CHARACTERSET=AL32UTF8
            - ORACLE_PWD=oracle
         build:
            context: .
            shm_size: '2gb'
         shm_size: '2gb'
         ports:
            - "1521:1521"
   
      kafka:
         image: "${KAFKA_IMAGE}"
         container_name: KAFKA1
         profiles: ["kafka"]
         volumes:
            - ./kafka:/var/lib/kafka/data
         networks:
            - isolated_network
         ports:
            - "9092:9092"
         environment:
         # KRaft parameters
            - CLUSTER_ID=${KAFKA_CLUSTER}
            - KAFKA_NODE_ID=1
            - KAFKA_PROCESS_ROLES=broker,controller
            - KAFKA_CONTROLLER_QUORUM_BOOTSTRAP_SERVERS=kafka:9093
         # Listeners
            - KAFKA_LISTENERS=PLAINTEXT://:9092,CONTROLLER://:9093
            - KAFKA_ADVERTISED_LISTENERS=PLAINTEXT://kafka:9092
            - KAFKA_CONTROLLER_LISTENER_NAMES=CONTROLLER
            - KAFKA_LISTENER_SECURITY_PROTOCOL_MAP=PLAINTEXT:PLAINTEXT,CONTROLLER:PLAINTEXT
         # Performance & configuration
            - KAFKA_GROUP_INITIAL_REBALANCE_DELAY_MS=0
            - KAFKA_OFFSETS_TOPIC_REPLICATION_FACTOR=1
            - KAFKA_TRANSACTION_STATE_LOG_REPLICATION_FACTOR=1
            - KAFKA_TRANSACTION_STATE_LOG_MIN_ISR=1
            - KAFKA_AUTO_CREATE_TOPICS_ENABLE=true
         healthcheck:
            test: ["CMD", "/kafka/bin/kafka-topics.sh", "--bootstrap-server", "${KAFKA_BROKER}", "--list"]
            interval: 10s
            timeout: 5s
            retries: 5
            start_period: 60s
   
      openlogreplicator:
         image: "${OLR_IMAGE}"
         container_name: "${OLR_CONTAINER}"
         profiles: ["openlogreplicator"]
         depends_on:
            - oracle
            - kafka
         volumes:
            - ./bin:/opt/bin
            - ./checkpoint:/opt/OpenLogReplicator/checkpoint
            - ./fra:/opt/fra
            - ./log:/opt/OpenLogReplicator/log
            - ./oradata:/opt/oradata
            - ./scripts:/opt/OpenLogReplicator/scripts
         restart: "no"
         ports:
            - "50000:50000"
         networks:
            - isolated_network
   
      kafka-connect:
         image: quay.io/debezium/connect:latest
         container_name: KAFKA-CONNECT
         profiles: ["connect"]
         environment:
            BOOTSTRAP_SERVERS: kafka:9092
            GROUP_ID: 1
            CONFIG_STORAGE_TOPIC: connect_configs
            OFFSET_STORAGE_TOPIC: connect_offsets
            STATUS_STORAGE_TOPIC: connect_statuses
         ports:
            - "8083:8083"
         depends_on:
            - oracle
            - kafka
         networks:
            - isolated_network
   
   networks:
      isolated_network:

   cd ../tutorials/Oracle-to-Kafka/
   ./2.db-start.sh

4. Create directory kafka:

   mkdir kafka
   chmod 777 kafka

   5. Add line to sql/schema-usrolr.sql:

   GRANT SELECT ON SYS.V_$PDBS TO USROLR;

5. Configure OpenLogReplicator:

   {
      "version": "1.9.0",
      "log-level": 4,
      "source": [
         {
            "alias": "S1",
            "name": "ORA1",
            "reader": {
            "type": "online",
            "path-mapping": ["/opt/oracle/oradata", "/opt/oradata", "/opt/oracle/fra", "/opt/fra"],
            "user": "USROLR",
            "password": "USROLRPWD",
            "server": "//oracle:1521/XEPDB1"
            },
            "format": {
               "type": "json",
               "db": 3,
               "interval-dts": 9,
               "interval-ytm": 4,
               "message": 2,
               "rid": 1,
               "column": 2,
               "schema": 7,
               "scn-type": 5,
               "timestamp-type": 4,
               "flush-buffer": 0
            },
            "filter": {
            "table": [
               {"owner": "USRTBL", "table": "TAB1.*"}
            ]
            }
         }
      ],
      "target": [
         {
            "alias": "T1",
            "source": "S1",
            "writer": {
            "type": "network",
            "uri": "0.0.0.0:50000",
            "queue-size": 1000000
            }
         }
      ]
   }

6. Modify file and run file 4.olr-start.sh:

   set -e
   
   . cfg.sh
   . ../common/functions.sh
   
   echo "4. creating and starting OpenLogReplicator container"
   olr_files
   db_sql "${DB_CONTAINER}" /opt/sql/schema-usrolr.sql /opt/sql/schema-usrolr.out
   docker_up --profile openlogreplicator --profile kafka --profile connect
   finish

   ./4.olr-start.sh

7. Create debezium connector:

   curl -X POST http://localhost:8083/connectors/ \
   -H "Content-Type: application/json" \
   -d '{
      "name": "oracle-olr-connector",
      "config": {
         "connector.class": "io.debezium.connector.oracle.OracleConnector",
   
         "database.hostname": "oracle",
         "database.port": "1521",
         "database.user": "sys as sysdba",
         "database.password": "oracle",
         "database.dbname": "XEPDB1",
   
         "database.connection.adapter": "olr",
   
         "openlogreplicator.host": "openlogreplicator",
         "openlogreplicator.port": "50000",
         "openlogreplicator.source": "ORA1",
   
         "topic.prefix": "oracle_olr",
         "snapshot.mode": "no_data",
   
         "schema.history.internal.kafka.bootstrap.servers": "kafka:9092",
         "schema.history.internal.kafka.topic": "schema-changes.oracle",
   
         "table.include.list": "USRTBL.TAB1",
         "database.history.store.only.captured.tables.ddl": "true"
      }
   }'

8. Start:

   CREATE TABLE USRTBL.TAB1(
      ID         NUMBER NOT NULL,
      COUNT      NUMBER
   );
   ALTER TABLE USRTBL.TAB1 ADD CONSTRAINT ID_PK PRIMARY KEY(ID);
   ALTER TABLE USRTBL.TAB1 ADD SUPPLEMENTAL LOG DATA (PRIMARY KEY) COLUMNS;
   
   INSERT INTO USRTBL.TAB1 (ID, COUNT)
   SELECT 
      LEVEL - 1 AS ID,
      MOD(LEVEL, 100) AS COUNT
   FROM DUAL
   CONNECT BY LEVEL <= 500000;
   
   COMMIT;

9. Restart OLR docker container during transaction:

docker restart OLR1

Restarting OLR will create Connection lost error in connector, so it need to be restarted too:

curl -X PUT http://localhost:8083/connectors/oracle-olr-connector/stop

curl -X PUT http://localhost:8083/connectors/oracle-olr-connector/resume

11. Observe results:

docker exec -it KAFKA1 /kafka/bin/kafka-get-offsets.sh \
--bootstrap-server localhost:9092 \
--topic oracle_olr.USRTBL.TAB1 \
--time -1

Actual result:

oracle_olr.USRTBL.ADAM1:0:499999

Expected result:

oracle_olr.USRTBL.ADAM1:0:500000

OLR logs with STREAM trace snippet after restart:

2026-04-21 11:02:23 INFO  00000 new client connection: 172.19.0.5:39222
2026-04-21 11:02:23 TRACE STRM  request: INFO: ORA1
2026-04-21 11:02:23 TRACE STRM  request: CONTINUE database: ORA1, c_scn: 2629475, c_idx: 28763
2026-04-21 11:02:23 INFO  00000 client requested scn: 2629475, idx: 28763
2026-04-21 11:02:23 INFO  00000 streaming to client
2026-04-21 11:02:24 TRACE STRM  data[421]: [{"scn":2631626,"tm":1776769311000000000,"c_scn":2629475,"c_idx":28765,"xid":"0x000a.01a.00000268","db":"XEPDB1","payload":[{"op":"c","schema":{"owner":"USRTBL","table":"TAB1","obj":75754,"columns":[{"name":"ID","type":"number","precision":-1,"scale":-1,"nullable":false},{"name":"COUNT","type":"number","precision":-1,"scale":-1,"nullable":true}]},"num":28763,"rid":"AAASftAANAAAAEEAAe","after":{"ID":28763,"COUNT":64}}]}]

In this case, OLR is expected to start streaming data from c_idx = 28764, but it starts from 28765 instead.

---

Proposed Fix

I noticed that in Builder.h, in the builderBegin() method, the parameter msg->lwnIdx is assigned before the lwnIdx variable is incremented:

msg->lwnIdx = lwnIdx++;

I tested changing it to:

msg->lwnIdx = ++lwnIdx;

After applying this change, the described issue no longer occurs.

I ran multiple test transactions, including long-running workloads, and did not observe any data loss. All messages were delivered correctly.

---

Environment details (required)

• Official binary signature/version: v1.9.0
• OpenLogReplicator version: 1.9.0
• Oracle version and edition: Oracle Database 21c Express Edition Release 21.0.0.0.0 - Production / Oracle Database 19c Standard Edition 2 Release 19.0.0.0.0 - Production
• Operating system: Debian 13.0 / Red Hat 9.7
• Docker version: 29.2.1

---