Combined Database
When you query combined database, actually the query goes to the archive and real time databases separately and then the results are joined together.
That's why this is the combination of features of these databases.
Select Block attributes for combined database controls how you can query the trunk or branch block updates ONLY for real time database data part.
Typically you should avoid using this type of query, as it is slower than real time and archive and does not give full consistency of the data.
Also Check Archive and RealTime dataset.
What is the difference between realtime, archive, and combined datasets?
Realtime holds a rolling recent window (roughly the last hours) with low latency and may include branch blocks. Archive holds genesis-to-near-present data with trunk consistency and higher ingest delay. Combined runs your query against both and merges results so one query can span recent + historical, but it is slower and can expose different fields than realtime-only (especially on Solana). Read Realtime, Archive, and Dataset options.
Why does data from archive dataset not match combined dataset?
Archive and realtime are separate pipelines: block inclusion, finality, and delay differ. Combined joins the two, so row counts, last timestamps, and aggregates may not equal archive alone plus realtime alone.
Why does dataset: combined return fewer fields than dataset: realtime on Solana?
On Solana, some projections (for example certain Trade.Side / account-level fields) are only populated in the realtime slice. Combined and archive historical aggregates may omit those columns, which triggers “columns not available” errors if you request them. Use fields documented for archive/combined (e.g. DEXTradeByTokens aggregates) or switch to dataset: realtime for debugging. See Historical Solana aggregate data and Pump.fun combined-dataset note.