1. Have a PostgeSQL instance with wal2Json plugin (I used PostgreSQL 10.3) 2. Have a PostgreSQL connector stood up, my config: { "connector.class": "io.debezium.connector.postgresql.PostgresConnector", "plugin.name": "wal2json", "tasks.max": "1", "database.hostname": "**", "database.port": "5432", "database.user": "**", "database.password": "**", "database.dbname": "postgres", "database.server.name": "elasticsearch", "table.whitelist": "public.dbo.testmssqltable", "decimal.handling.mode": "double" } 3. Create a source table with column of timestamp data type. 4. Insert record with timestamp value that has NanoOfSecond and has minus sign, e.g.: 1936-10-25 22:10:12.608 5. Error is thrown e.g.: [2018-05-10 08:45:45,860] ERROR Failed to properly convert data value for 'public.dbo.testmssqltable.DateTime_Test' of type timestamp for row [..., 1918-08-20 02:30:08.123, ...] : (io.debezium.relational.TableSchemaBuilder) kafka-connect_1 | java.time.DateTimeException: Invalid value for NanoOfSecond (valid values 0 - 999999999): -877000000 kafka-connect_1 | at java.time.temporal.ValueRange.checkValidValue(ValueRange.java:311) kafka-connect_1 | at java.time.temporal.ChronoField.checkValidValue(ChronoField.java:703) kafka-connect_1 | at java.time.LocalTime.of(LocalTime.java:342) kafka-connect_1 | at java.time.LocalDateTime.of(LocalDateTime.java:362) kafka-connect_1 | at io.debezium.time.Conversions.toLocalDateTime(Conversions.java:140) kafka-connect_1 | at io.debezium.time.NanoTimestamp.toEpochNanos(NanoTimestamp.java:68) kafka-connect_1 | at io.debezium.jdbc.JdbcValueConverters.convertTimestampToEpochNanos(JdbcValueConverters.java:496) kafka-connect_1 | at io.debezium.connector.postgresql.PostgresValueConverter.convertTimestampToEpochNanos(PostgresValueConverter.java:402) kafka-connect_1 | at io.debezium.jdbc.JdbcValueConverters.lambda$converter$24(JdbcValueConverters.java:320) kafka-connect_1 | at io.debezium.relational.TableSchemaBuilder.lambda$createValueGenerator$2(TableSchemaBuilder.java:197) kafka-connect_1 | at io.debezium.relational.TableSchema.valueFromColumnData(TableSchema.java:111) kafka-connect_1 | at io.debezium.connector.postgresql.RecordsSnapshotProducer.generateReadRecord(RecordsSnapshotProducer.java:311) kafka-connect_1 | at io.debezium.connector.postgresql.RecordsSnapshotProducer.readTable(RecordsSnapshotProducer.java:270) kafka-connect_1 | at io.debezium.connector.postgresql.RecordsSnapshotProducer.lambda$takeSnapshot$6(RecordsSnapshotProducer.java:195) kafka-connect_1 | at io.debezium.jdbc.JdbcConnection.queryWithBlockingConsumer(JdbcConnection.java:407) kafka-connect_1 | at io.debezium.connector.postgresql.RecordsSnapshotProducer.takeSnapshot(RecordsSnapshotProducer.java:193) kafka-connect_1 | at io.debezium.connector.postgresql.RecordsSnapshotProducer.lambda$start$0(RecordsSnapshotProducer.java:80) kafka-connect_1 | at java.util.concurrent.CompletableFuture$AsyncRun.run(CompletableFuture.java:1626) kafka-connect_1 | at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142) kafka-connect_1 | at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617) kafka-connect_1 | at java.lang.Thread.run(Thread.java:745) You can move points 3 and 4, before starting connector - I observe same behaviour. All records that follow looks like are streamed correctly. If initial records or records inserted in first batch doesn't have any such vale (minus timestamp), Error is not present and everything is streamed correctly.