Command failed with error 286 (ChangeStreamHistoryLost) / error 280 (ChangeStreamFatalError): resume point was not found in the oplog

Root cause

The MongoDB resume token stored in the connector's offset refers to a position that has rolled off the oplog. The oplog is a capped collection — old entries are removed as new ones arrive. If the connector is stopped for longer than the oplog window (typically 24–72h on Atlas), the token becomes invalid.

How to fix

1. Increase oplog size — mongod.conf: storage.oplogSizeMB: 51200 (50 GB). Minimum: accommodate your peak write rate × 2 hours of headroom.
2. Enable heartbeat.interval.ms in Debezium to keep the resume token fresh even during low-activity periods.
3. To recover: delete the connector, clear its offset from the Kafka Connect offsets topic, and recreate it (full re-snapshot).
4. On MongoDB Atlas: increase oplog retention in the cluster configuration.