If Dgraph uses the bulk loader, does it mean that scaling is not possible? Based on my understanding of the code, the w directory is used to record write-ahead logs (WAL). When a new node joins, synchronization is performed through sequential writes based on snapshots in the w directory. However, in bulk load mode, data is not written to the w directory. Does this mean that the initial data loaded via bulk load cannot be retrieved by new nodes?
That’s correct. Bulk loading is intended for the initial creation of the graph, not batch updating. Check out the “live loader” for that use case.