Snowflake runs the SQL. Kestra decides when it runs, what feeds it, and what runs next. Load stages on file arrival, run idempotent MERGE upserts as transactions, chain with dbt and ingestion, and trace every statement in one execution history.
Snowflake Tasks and Streams cover scheduling inside the warehouse. Kestra covers everything around it: it loads files to a stage and runs COPY the moment they land, chains Snowflake with Airbyte ingestion and dbt transforms in one flow, and runs incremental MERGE upserts as transactions so a partial load never lands. A transient dbt retry never re-runs the load, and every query is tagged back to the run that issued it.
Snowflake runs the SQL, the warehouse, and the stage. Kestra runs the steps around the query: what triggers it, what feeds it, who approves it, and where the cross-tool audit trail lives.
Snowflake Tasks fire on a cron schedule. If the upstream load slips, the task runs on stale data. Kestra triggers io.kestra.plugin.jdbc.snowflake.Query the moment the upstream step confirms, whether that is an Airbyte sync, a file landing in S3, or a webhook. The Trigger task can also poll Snowflake and start a flow when a query returns rows.
Loading data means PUT a file to a stage, then COPY INTO a table, then validate. Kestra runs Upload to push a file from internal storage to a Snowflake stage, then Query to run the COPY, then a row-count check, as ordered steps with per-step retries. A failed COPY retries without re-uploading the file.
Snowflake has no view of the connector that fed it or the dbt models that read it. Kestra runs ingestion, the Snowflake load, the transform, and the Hightouch activation as steps in one flow. Run IDs and the Snowflake query ID flow forward, and the whole chain shares one execution ID.
An incremental load is rarely one statement: stage the file, MERGE into the target, dedup, update a watermark. io.kestra.plugin.jdbc.snowflake.Queries runs them as one transaction (enabled by default), so a failure rolls back and a partial load never lands. The next run picks up cleanly from the last committed watermark.
Idle warehouses burn credits. Kestra can resume a warehouse with a query, run the workload, and suspend it again in the same flow, so compute is only on for the job. Query tags set per run tie Snowflake's query history and credit usage back to the exact Kestra execution that issued each statement.
Analysts should not need warehouse access to run a parameterized export. Kestra Apps wraps a Query flow in a typed form: pick the date range, run the query, get the result file. Every run lands in execution history with the requesting user, so self-service does not mean unaudited.
Patterns data platform teams run in production today. Each one shows the flow end to end, with the real plugin classes in play.
Airbyte syncs raw data into Snowflake, Kestra waits for it, then runs the Snowflake load and triggers dbt against the warehouse. Slack confirms. A dbt retry never re-runs the Airbyte sync.
A dbt retry never re-runs Airbyte; a load retry never re-runs ingestion.
The Snowflake query ID lands in outputs for downstream traceability.
Debug ingestion, load, and transform from one screen, not three.
A file lands in object storage. Kestra uploads it to a Snowflake stage with Upload, runs Query for the COPY INTO, then checks the loaded row count. A failed COPY retries without re-uploading the file.
Upload pushes the file; Query runs the COPY only after it lands.
An If task branches on the loaded row count before anything reads the table.
A COPY failure retries without re-staging the file.
Some pipelines react to data, not clocks. The Snowflake Trigger polls the warehouse on an interval and starts a flow when a sentinel query returns rows, then marks them processed with afterSQL so the same rows do not fire twice.
Fires on rows returned, not on the next cron tick.
Sentinel rows are flagged so they do not re-trigger.
A ForEach processes each returned row independently.
The daily pattern: COPY new rows into a staging table, then MERGE into the target with dedup and a watermark update, all as one transaction with Queries. A failed MERGE rolls back, so the target never holds a half-applied batch.
Queries wraps the MERGE and watermark update; a failure rolls back.
The watermark means a rerun resumes from the last committed batch.
Late or duplicate rows collapse on the merge key, not downstream.
Kestra resumes a warehouse, runs the batch, and suspends it again in one flow, so compute is only billed for the job. A query tag set per run ties Snowflake credit usage back to the exact execution, and a scheduled flow can report daily spend to Slack.
Resume and suspend bracket the workload in the same flow.
Snowflake credit usage ties back to the issuing execution.
Queries runs several statements in one transactional task.
| Capability | | | | |
|---|---|---|---|---|
| Trigger Snowflake on upstream completion | Native event triggers | Cron or stream-based only | Sensor-based polling | Sensors, Python required |
| Stage upload plus COPY in one flow | Upload + Query tasks | SQL only, no file staging | Operator chain | Ops, Python required |
| Transactional multi-statement load | Queries in one transaction | Stored procedures | Operator chain | Ops, Python required |
| Trigger a flow on query results | Snowflake Trigger task | Streams + Tasks, in-warehouse | Custom sensor | Custom sensor |
| Chain with Airbyte, dbt, Hightouch | Native plugins | Snowflake-only | Operator-dependent | Asset-based, Python |
| Self-service queries for analysts | Kestra Apps | No | No native form layer | No native form layer |
| Declarative YAML, IaC-friendly | YAML + Terraform provider | SQL DDL | Python DAGs | Python assets |
Find answers to your questions right here, and don't hesitate to Contact Us if you couldn't find what you're looking for.
Beyond Snowpipe: Use Kestra for Complete Snowflake Workflow Orchestration
Kestra & Snowflake Partnership: How our users orchestate their Data Warehouse
Kestra Becomes a Registered Partner with Snowflake for Enhanced Data Workflow Management
Snowflake data pipeline with Kestra
Trigger on upstream events, load stages, run idempotent MERGE upserts, and chain with Airbyte and dbt in one flow. Open source, self-hosted, event-driven.
