Stop padding cron buffers. Trigger dbt Cloud the moment Fivetran completes, S3 files land, or Kafka events fire. Coordinate Hightouch downstream, gate production behind manual approval, and span every step in one execution history.
dbt Cloud solves transformation scheduling inside the dbt layer. Kestra solves what surrounds it: when jobs trigger, what runs after them, and where the audit trail lives.
dbt Cloud fires on a cron schedule or a Git event. If your Fivetran sync runs long, dbt starts building on stale data. Kestra triggers the dbt Cloud job the moment the upstream step confirms completion, whether that's a Fivetran sync, a file arrival in S3, or a webhook from any API. It halts the downstream activation layer until dbt confirms success.
dbt Cloud's lineage stops at the dbt boundary. It has no visibility into the Fivetran connector feeding it or the Hightouch sync consuming the output. Kestra runs all three as steps in a single flow. If dbt fails, Kestra retries only the dbt step. Fivetran does not re-sync. Outputs flow forward between steps, and the entire pipeline shares one execution ID.
Something breaks in the activation layer. You need to know whether dbt completed cleanly, which run produced the data, and who triggered it. dbt Cloud's history covers the dbt job. It has no record of the Fivetran sync that preceded it or the Hightouch sync that followed. Kestra logs every step across every tool with a trigger source, a timestamp, and a direct link to the failing task.
Patterns data engineering teams run in production. Pick one, copy the YAML, ship it through CI.
Fivetran syncs raw data, then Kestra triggers the dbt Cloud job and waits for it to confirm success, then fires the Hightouch sync. All three run as sequential steps in a single Kestra flow. If dbt fails, Kestra retries it without re-running the Fivetran sync.
Retry only the failed step. Fivetran does not re-sync on a dbt failure.
Pass artifacts and run IDs between Fivetran, dbt Cloud, and Hightouch.
Single place to debug the full pipeline, not three separate dashboards.
Replace the cron schedule that fires whether source data is ready or not. Kestra listens for the Fivetran sync completion event and triggers the dbt Cloud job the moment ingestion confirms success. No polling, no builds starting before the source data is confirmed ready.
dbt Cloud fires when the data is actually ready, not on the next interval.
Fire from any HTTP call, Kafka message, S3 event, or database change.
Kestra waits for each step to confirm before starting the next.
Run the dbt Cloud job against the staging environment. Pause the flow on the output. A named reviewer reads the result in the Kestra UI and resumes the flow. Only then does the production job trigger. The reviewer's identity and timestamp are logged automatically in the execution history.
Built-in Pause task suspends execution until a reviewer approves.
Identity and approval moment stored in execution history automatically.
Trigger different dbt Cloud job IDs for staging and production in the same flow.
Every production promotion is linked to a reviewer and a moment in time.
Some pipelines require multiple dbt Cloud jobs in a specific sequence: ingestion models first, then downstream marts, then final activation models. Kestra runs each job in order, waits for confirmation, and passes the run ID forward. If job 2 fails, job 3 does not start. A single Slack alert links directly to the failed step.
Each dbt Cloud job runs only after the previous confirms success.
Reference the upstream job's run ID in downstream steps for traceability.
One Slack message with a link to the exact job that failed.
Trigger different dbt Cloud jobs from the same Kestra flow by passing the job ID and custom environment variables as flow inputs at run time. One flow definition covers both the nightly full build job and the on-demand incremental job. The trigger or input determines which job fires, no duplicate flow definitions required.
Pass job IDs, environment names, and custom fields as typed inputs.
Nightly full builds and on-demand refreshes share one YAML.
Non-engineers trigger specific jobs through a form without touching YAML.
We switched from orchestrating with dbt Cloud because Kestra hosted Core really well.
Production-ready workflows. Copy the YAML, wire it to your stack.
Fivetran syncs raw data on a schedule. When it completes, Kestra triggers the dbt Cloud job and waits for success. On completion, it fires the Hightouch sync. Any failure stops the pipeline at the failing step and posts a Slack alert with a link to the execution.
Triggers the staging dbt Cloud job and pauses the flow on the result. A named reviewer reads the output in the Kestra UI and resumes the flow. Only then does the production job run. The reviewer's identity and timestamp are stored automatically in the execution history.
Runs three dbt Cloud jobs in order: ingestion models, mart models, activation models. Each step waits for the previous to confirm success before starting. If any job fails, Kestra stops the chain and posts a Slack alert with the run ID of the failing job and a link to the execution.
| Capability | | | GitHub Actions | |
|---|---|---|---|---|
| Trigger dbt Cloud on upstream tool completion | Native event triggers | Event polling, Python code required | Cron or Git events only | Sensor-based polling |
| Coordinate ingestion and activation around dbt | Native, with outputs | Possible, Python code required | Possible, glue scripts required | Yes, with operators |
| Approval gate before production job | Native pause + resume | Available in Prefect 2.x+ | Environment approvals (deploy only) | Custom sensor workaround |
| Retry failed step without re-running pipeline | Configurable per task | Task-level retry | Re-run from start | Task-level retry |
| Cross-stack execution history | One view across all tools | Flow run UI, Prefect-scoped | Actions UI, pipeline-scoped | DAG view, Airflow-scoped |
| Self-service for non-engineers | Kestra Apps | No native form layer | No | No |
| Chain dbt Cloud with Fivetran, Hightouch, Spark | 1300+ plugins | Possible, Python code required | Possible, fragile | Yes, operator-dependent |
Find answers to your questions right here, and don't hesitate to Contact Us if you couldn't find what you're looking for.
Trigger on upstream events, chain with Fivetran and Hightouch, and track every run across your full data stack. Open source, self-hosted, event-driven.
