Read first:
../conectores.md(state map) andfichas.md(index of sheets).
What it is#
Reads the tables of a Snowflake database and schema via the SQL API (paginated SQL queries, key-pair authentication) and loads them into the Bronze layer.
Supported ingestion modes:
- We pull — We pull the data from your source.
What it's for#
Bring Snowflake data into your datalake and make it available to the Bronze/Silver/Gold layers, BI models and the natural language question wizard.
When to use / when not to use#
Use when you need to consolidate this source into the datalake on a scheduled basis, with isolation per workspace.
Remember that the first run with a real credential is followed by our team before the connector is fully released.
Plan and permissions#
- Plan: core product (connectors are part of all plans; the limit is the number of sources in the plan).
- Access: requires an active workspace — if the subscription is overdue, suspended or canceled, the run is blocked.
- Disablement: the ingestion run can be temporarily disabled (maintenance or incident). When this happens, the screen indicates that the feature is unavailable at the moment.
Prerequisites#
- A Snowflake account with access to the data you want to ingest.
- Generate an RSA key pair (openssl genrsa 2048 | openssl pkcs8 -topk8 -nocrypt) and register the public key on the Snowflake user (
ALTER USER usuario SET RSA_PUBLIC_KEY='…'). Provide the account identifier (e.g.: xy12345.sa-east-1), the user, the database and the schema to read, and paste the PRIVATE key in PEM. It is encrypted (AES-256-GCM), never shown again and is only used to sign the short-lived access token.
Credentials#
Authentication method: RSA key pair: you register the public key on the warehouse user (in Snowflake, ALTER USER … SET RSA_PUBLIC_KEY) and paste the private key here (PKCS#8, no passphrase). The private key is encrypted (AES-256-GCM), never shown again and is only used to sign a short-lived access token (59 min) inside the call — it does not travel to the provider nor appear in logs/errors.
Sensitive fields. They are encrypted (AES-256-GCM) and never remain in the config of the connection, in the pipeline definition, in the export or in the DAG. By default the encrypted value is stored in the database; moving the credential of this connection to the external credential vault is an explicit choice made per connection after saving — see ../../administracao/credenciais-e-rotacao.md.
- Private key (PEM) (
privateKey) — RSA private key in PKCS#8 without passphrase, pair of the public key registered on the user.
Secrets are never re-displayed and are redacted in any error/log (raw token, URL-encoded and in JSON).
Form fields#
| Campo | Rótulo | Tipo | Obrigatório | Observação |
|---|---|---|---|---|
account | Account identifier | text | yes | The part before .snowflakecomputing.com in your account URL. |
usuario | User | text | yes | — |
banco | Database | text | yes | — |
schema | Schema | text | yes | default: PUBLIC; Only tables in this schema are listed. |
warehouse | Warehouse | text | no | Optional: warehouse used in queries (otherwise, the user's default). |
privateKey 🔒 | Private key (PEM) | password/secret | yes | secret (encrypted; optional external vault); RSA private key in PKCS#8 without passphrase, pair of the public key registered on the user. |
Step-by-step#
- In the Dados → Conexões menu, use Nova conexão.
- Choose Snowflake in the catalog.
- Fill in the form fields (see the table above).
- Use Testar conexão. The result is shown separated by Network (server reachability), Authentication (user and password) and Read (list tables) — each with OK, Failed, Not applicable or Not tested.
- Use Salvar conexão. Saving DOES NOT start any ingestion.
- On the connection card, use Montar extração (or go to Orquestração → Pipelines and use Conectar fonte). The pipeline points to the saved connection — the credential is not copied.
- In the review, Publicar (agenda desligada) creates version 1 without running anything. Afterwards, Ativar agenda and Executar agora are separate actions.
Example — connection test#
Example configuration (all values are fictitious; secrets appear masked):
# Exemplo sintético (valores fictícios — não são credenciais reais)
account = minhacontaexemplo
usuario = <usuario>
banco = <banco>
schema = public
warehouse = <warehouse>
privateKey = •••••• (cifrado — nunca exibido)Expected result#
The Testar conexão returns the list of discovered tables/resources. After the first load, the rows appear in the Bronze layer of your workspace, ready for the pipeline.
Limits and costs#
- Pagination: hard cap per run (default 10,000 pages / 5,000,000 rows); anti-loop guard.
- SaaS load: default
batchSize5,000,maxRows200,000 per resource. - Estimated volume: ~1 GB/month (catalog estimate, not a measurement).
- Cost: the extraction itself is not charged by the platform; the Bronze load consumes datalake processing (billable).
Security#
- Scope per workspace: each customer reads and writes only to their own datalake (
<prefixo>_bronze/_silver/_gold); there is never leakage between workspaces. - Encrypted secrets: sensitive material is encrypted with AES-256-GCM before being written. The external credential vault is optional and linked per connection; without it, the encrypted value is stored in the database. Changing the value here DOES NOT revoke the credential at the source.
- Redaction: all errors/logs go through secret redaction (raw, URL-encoded and in JSON).
- Production does not fake: a path that would fall into a mock triggers the production guard (explicit failure instead of simulating success).
Common errors#
- Credential rejected (401/403) — the provider did not accept the credential: expired, revoked or lacking required permission. Reconnect the source (the source and history are preserved; do not recreate the source).
- Rate limit (429) — the provider asked to wait. The extraction respects
Retry-Afterand retries; if it repeats, wait and reduce the schedule frequency. - Resource not found — check the resource identifier (account, object, report) and the granted permissions.
- Stuck pagination cursor — the extraction stops when the API cursor does not advance (anti-loop guard). If it happens in consecutive runs, open a support ticket with the connector and resource information.
- Empty records path — the response arrived with no records in the expected place. Check the selected resource and filters (period, account, status) before assuming there is no data.
Troubleshooting#
- Use Testar conexão to isolate credential/network problems before scheduling.
- Monitor runs in the observability per connector (error rate, p50/p95, freshness).
- A schema change at the source is signaled by schema drift (new column = info; removed column/type changed = attention).
Related#
../conectores.md— connector catalog by state../ingestao-incremental-backfill.md— full, incremental by watermark, backfill and checkpoint../pipeline-medalhao.md— Bronze/Silver/Gold layers and quality../agendamento-retry-dlq.md— scheduling, retry/backoff and DLQ../custo-limites-seguranca.md— cost, caps and workspace security
Last revision: 2026-10-04. Sheet automatically generated from the product connector catalog.