With the Authoritative Shards program, introduced in v9.0, a shard’s catalog may no longer contain chunks that the shard does not currently own. However, to serve snapshot reads, a shard must retain information about chunks it owned in the past when that historical information is still needed to target point-in-time reads.

Before v9.0, each shard’s catalog contained all chunks for a sharded collection. As a result, there was no additional complexity associated with retaining historical ownership information in the shard catalog for point-in-time reads.

In v9.0, chunk migrations can remove this historical ownership information even when it is still required for point-in-time reads. To prevent that from happening, moveChunk and moveRange operations should fail with a ConflictingOperationInProgress error if the migration would remove ownership history that is still required.

The same command should succeed once the snapshot history window, defined by minSnapshotHistoryWindowInSeconds, has elapsed.

Important: Note that this error will never be hit in a cluster where orphanCleanupDelaySecs is higher than minSnapshotHistoryWindowInSeconds.