<!-- llms-explorer concept facts · https://llms-explorer.com/tree/atlas-log-export-replay-destination-changes-and-failure-state-observations/ · pack 2026-11-02 · ~1047 tokens -->

# Atlas log-export replay destination changes and failure-state observations

> Source-anchored research for Atlas log-export replay destination changes and failure-state observations. Run status: COMPLETED.

Parent: [Atlas External Log Sink to GCS](https://llms-explorer.com/tree/atlas-external-log-sink-to-gcs/) · 4 facets · 8 facts · page: https://llms-explorer.com/tree/atlas-log-export-replay-destination-changes-and-failure-state-observations/

## Definitions

- `LOG_STREAMING_EXPORT_RECOVERED` means queued logs are ready to resume sending. The event reference marks it non-alertable, and the log-export overview lists recovery readiness as an informational Project Activity Feed event. (medium confidence) (verified as of 2026-11-02) — [source](https://www.mongodb.com/docs/atlas/reference/atlas-alert-event-types.md) · [also](https://www.mongodb.com/docs/atlas/export-logs.md)
- `LOG_STREAMING_REPLAY_STARTED` means queued logs are now being sent. The event reference marks it non-alertable, and the log-export overview lists this sending state as an informational Project Activity Feed event. (medium confidence) (verified as of 2026-11-02) — [source](https://www.mongodb.com/docs/atlas/reference/atlas-alert-event-types.md) · [also](https://www.mongodb.com/docs/atlas/export-logs.md)
- `LOG_STREAMING_REPLAY_COMPLETE` means all queued logs have been delivered to the sink. The event reference marks it non-alertable, and the log-export overview lists delivery completion as an informational Project Activity Feed event. (medium confidence) (verified as of 2026-11-02) — [source](https://www.mongodb.com/docs/atlas/reference/atlas-alert-event-types.md) · [also](https://www.mongodb.com/docs/atlas/export-logs.md)

## How it works

- The retries-exhausted alert triggers after delivery retries are exhausted; its message says logs are being queued and will be sent once the connection is restored. The event reference identifies it as alertable `LOG_STREAMING_EXPORT_FAILED_RETRIES_EXHAUSTED`. (medium confidence) (verified as of 2026-11-02) — [source](https://www.mongodb.com/docs/atlas/reference/alert-conditions/) · [also](https://www.mongodb.com/docs/atlas/reference/atlas-alert-event-types.md)

## Problems, failure modes and limitations

- The non-retryable failure alert triggers immediately for an error such as invalid credentials or a misconfigured sink. The event reference identifies this alertable condition as `LOG_STREAMING_EXPORT_FAILED_NONRETRYABLE`. (medium confidence) (verified as of 2026-11-02) — [source](https://www.mongodb.com/docs/atlas/reference/alert-conditions/) · [also](https://www.mongodb.com/docs/atlas/reference/atlas-alert-event-types.md)
- The replay-failure alert concerns automatic replay from the DLQ failing after an automatic retry attempt. The alert-conditions documentation tells users to contact MongoDB Support to request manual redelivery and warns that undelivered records will be permanently deleted when their retention window expires. (medium confidence) (verified as of 2026-11-02) — [source](https://www.mongodb.com/docs/atlas/reference/alert-conditions/) · [also](https://www.mongodb.com/docs/atlas/export-logs.md)
- For Activity Feed event export, the pinned Delivery Behavior section explicitly labels the initial release as having no automatic replay, backfill or failure alerts. It says persistent failures do not automatically replay events, historical events from before integration configuration are not backfilled, and users should contact MongoDB Support to request redelivery after a persistent failure. This scopes the system-log replay behavior separately from event export. (low confidence) (verified as of 2026-11-02) — [source](https://www.mongodb.com/docs/atlas/export-events.md)

## Comparisons and alternatives

- In the pinned Atlas system-log documentation, modifying the sink configuration while logs are queued in the DLQ makes replay use the new configuration. The documented example changes the destination between S3 buckets: queued logs go to the new bucket rather than the original destination. (low confidence) (verified as of 2026-11-02) — [source](https://www.mongodb.com/docs/atlas/export-logs.md)
