Loading, updating, deleting, and exporting have different correctness and throughput constraints. Start from the desired state change.
Doris supports synchronous HTTP loads, continuous Kafka ingestion, SQL-based ingestion, and asynchronous batch workflows.
Synchronous files or streams with immediate job results.
Continuous Kafka consumption with managed progress.
Transform and load from tables, Catalogs, or file TVFs.
Small SQL writes where simplicity matters more than bulk throughput.
Combine high-frequency small writes into efficient commits.
Track larger asynchronous jobs by label and state.
Labels, quality thresholds, error samples, job state, and source offsets turn ingestion from a request into an operated workflow.
Unique Key tables support UPSERT behavior and SQL UPDATE; Aggregate Key tables update through their declared aggregation functions.
Write the complete new logical row.
Write only selected fields of a wide row.
Predicate DELETE handles supported row filters, load-based delete signs handle batches, and TRUNCATE clears a table or partition quickly.
Remove supported rows matching a bounded condition.
Apply a large set of keyed deletions through ingestion.
Clear a full table or partition without row filtering.
SELECT INTO OUTFILE exports transformed query results, EXPORT runs larger asynchronous extraction jobs, and mysqldump captures schema or small logical datasets.
Synchronous, transformable results in CSV, Parquet, or ORC.
Asynchronous extraction for larger table or partition scopes.
Schema export and small logical backups through MySQL tooling.
A reliable operation has an explicit scope, correctness boundary, observable result, retry strategy, and downstream verification.
