# MongoServerError: ChangeStream error: resume token from WiredTiger is stale: token timestamp 1234567890 is older than oldest timestamp 1234567895

- **ID:** `mongodb/change-stream-resume-token-wiredtiger-stale`
- **Domain:** mongodb
- **Category:** runtime_error
- **Verification:** ai_generated
- **Fix Rate:** 75%

## Root Cause

The resume token timestamp has been pruned from WiredTiger's history because the storage engine's oldest timestamp advanced past it due to checkpoint or replication pressure.

## Version Compatibility

| Version | Status | Introduced | Deprecated |
|---------|--------|------------|------------|
| mongodb 6.0 | active | — | — |
| mongodb 7.0 | active | — | — |
| mongodb 8.0 | active | — | — |

## Workarounds

1. **Increase the oplog size to retain more history: `db.adminCommand({ replSetResizeOplog: 1, size: 40960 })` (40GB) and use `db.adminCommand({ setParameter: 1, minSnapshotHistoryWindowInSeconds: 3600 })` to keep WiredTiger snapshots for 1 hour.** (85% success)
   ```
   Increase the oplog size to retain more history: `db.adminCommand({ replSetResizeOplog: 1, size: 40960 })` (40GB) and use `db.adminCommand({ setParameter: 1, minSnapshotHistoryWindowInSeconds: 3600 })` to keep WiredTiger snapshots for 1 hour.
   ```
2. **Implement fallback logic in the application: on this error, start a new change stream from the current time using `db.collection.watch([], { startAfter: resumeToken })` or `{ startAtOperationTime: Timestamp(...) }` and reprocess recent events.** (90% success)
   ```
   Implement fallback logic in the application: on this error, start a new change stream from the current time using `db.collection.watch([], { startAfter: resumeToken })` or `{ startAtOperationTime: Timestamp(...) }` and reprocess recent events.
   ```
3. **Reduce the resume token interval by polling change streams more frequently (e.g., every 5 seconds instead of 30 seconds) to keep tokens within the retention window.** (70% success)
   ```
   Reduce the resume token interval by polling change streams more frequently (e.g., every 5 seconds instead of 30 seconds) to keep tokens within the retention window.
   ```

## Dead Ends

- **** — The token is already pruned from the oplog and WiredTiger; retrying with the same token will result in the same error. (95% fail)
- **** — Checkpoint frequency does not directly control timestamp pruning; the oldest timestamp is managed by replication and read concern settings. (80% fail)
- **** — This is not configurable and would break point-in-time read operations; the error is a symptom of a broader retention issue. (90% fail)