Data Loading, Unloading, and Connectivity · Topic 6 of 7
Snowflake Drivers and Connectors
Know the Snowflake drivers (JDBC, ODBC, Python, Go, .NET, Node.js, PHP) and connectors for Kafka and Spark, plus the SQL API and native connectors.
- 5 min
- Drivers: JDBC, ODBC, Python connector, Go, .NET, Node.js, PHP PDO.
- SQL API = REST over HTTPS for SQL statements, no driver.
- Kafka connector runs in Kafka Connect and loads through Snowpipe or Snowpipe Streaming.
What it is
Applications and tools reach Snowflake through drivers and connectors. A driver lets a program written in a given language connect, run SQL and fetch results. A connector moves data between Snowflake and another platform, such as Apache Kafka or Apache Spark, or brings in data from a SaaS application.
The exam asks you to match a tool or language to the right option, and to know which connector loads data in which way.
How it works
Drivers
| Driver | Language or use | Typical client |
|---|---|---|
| JDBC Driver | Java | Java applications, many BI and ETL tools |
| ODBC Driver | Any ODBC client | Excel, BI tools, C and C++ programs |
| Snowflake Connector for Python | Python (implements the Python DB API) | Scripts, notebooks, Airflow jobs |
| Go Snowflake Driver | Go | Go services |
| .NET Driver | C# and other .NET languages | Windows and .NET applications |
| Node.js Driver | JavaScript on Node.js | Web back ends |
| PHP PDO Driver for Snowflake | PHP | PHP web applications |
All drivers authenticate the same way Snowflake does elsewhere: password with MFA, key pair authentication, OAuth, or external browser single sign-on, depending on the driver and account settings. Every query still runs on a virtual warehouse chosen by the session.
The Snowflake Connector for Python is named a "connector" but sits in the drivers family. Questions that ask how a Python script runs SQL against Snowflake point to it; questions about DataFrame code that runs inside Snowflake point to Snowpark.
Other ways to connect
- SQL API: a REST API that submits SQL statements over HTTPS and returns results as JSON, with no driver to install. It suits serverless functions and languages without a driver.
- Snowflake CLI: the command line tool for running SQL and managing objects and apps from a terminal or CI pipeline.
- Snowpark: libraries for Python, Java and Scala that build DataFrame operations which execute inside Snowflake.
- Snowsight: the browser interface, which needs no driver at all.
Connectors
| Connector | What it does | Load mechanism |
|---|---|---|
| Snowflake Connector for Kafka | Reads Kafka topics and writes them into Snowflake tables | Snowpipe (files buffered from topic messages) or Snowpipe Streaming (rows, lower latency) |
| Snowflake Connector for Spark | Lets Spark read Snowflake tables as DataFrames and write DataFrames back | Uses Snowflake stages and COPY under the hood, with query pushdown to Snowflake |
| Native connectors | Snowflake built connectors for SaaS applications and databases, installed as Snowflake Native Apps from Snowflake Marketplace | Pull data from the source into tables on a schedule |
| Partner tools | ETL, ELT and BI tools from Snowflake partners, many available through Partner Connect | Each tool's own method, usually through a driver |
The Kafka connector in more detail
The Kafka connector runs in a Kafka Connect cluster, outside Snowflake. Each topic maps to a table. With the Snowpipe ingestion method, it writes topic messages to files in an internal stage and calls Snowpipe to load them. With Snowpipe Streaming, it sends rows directly through channels, with no staged files.
By default each row holds two VARIANT columns: RECORD_CONTENT (the message) and RECORD_METADATA (topic, partition, offset and other details). Schema detection and evolution, when enabled, map message fields to separate columns instead.
The Spark connector in more detail
The Spark connector supports query pushdown: filters, projections and aggregates in Spark code are translated into SQL and run in Snowflake, so less data moves to the Spark cluster. It is the choice when an existing Spark pipeline, for example on Databricks or Amazon EMR, must read from or write to Snowflake.
Do not confuse the Spark connector with Snowpark. The Spark connector runs your code on a Spark cluster and exchanges data with Snowflake. Snowpark runs DataFrame code inside Snowflake on a virtual warehouse, with no Spark cluster.
Picking the right option
| Scenario | Answer |
|---|---|
| A BI tool needs a live connection | ODBC or JDBC driver, depending on the tool |
| A Python script loads a CSV and runs SQL | Snowflake Connector for Python (PUT and COPY INTO) |
| An app with no driver must run SQL over HTTPS | SQL API |
| Kafka events must land in Snowflake within seconds | Kafka connector with Snowpipe Streaming |
| A Spark job must read Snowflake data with pushdown | Snowflake Connector for Spark |
| SaaS application data must be loaded with minimal code | A native connector from Snowflake Marketplace or a partner ETL tool |
Common mistakes
Picking Snowpark for a Spark cluster. Snowpark is not a Spark connector; it runs inside Snowflake.
Thinking connectors use their own compute. Data loaded by the Kafka connector uses serverless Snowpipe or Snowpipe Streaming. Queries pushed down by the Spark connector run on the warehouse of the Spark connection.
Treating the SQL API as a driver. It is a REST API with no library to install, not one of the language drivers.
Assuming the Kafka connector creates flat columns by default. Without schema detection it loads RECORD_CONTENT and RECORD_METADATA VARIANT columns.
What to remember for the exam
- Drivers: JDBC, ODBC, Python connector, Go, .NET, Node.js, PHP PDO.
- SQL API = REST over HTTPS for SQL statements, no driver.
- Kafka connector runs in Kafka Connect and loads through Snowpipe or Snowpipe Streaming.
- Spark connector exchanges DataFrames with Snowflake and pushes queries down to Snowflake.
- Native connectors install as Snowflake Native Apps and pull SaaS data into tables.
Check yourself
Which driver does a Java application use to connect to Snowflake?
The JDBC Driver.
What two columns does the Kafka connector create in a target table by default?
RECORD_CONTENT for the message and RECORD_METADATA for topic, partition, offset and other details.
What does query pushdown in the Spark connector do?
It translates Spark operations into SQL that runs in Snowflake, so less data is moved to the Spark cluster.
Topic 6 of 7 in Data Loading, Unloading, and Connectivity
Test yourself on Data Loading, Unloading, and Connectivity
The full Snowflake SnowPro Core COF-C03 question bank has 845 questions, with timed practice exams and an explanation for every answer.
This study guide is independent and is not affiliated with, endorsed by, or sponsored by Snowflake. Vendor names, certification names and exam codes are trademarks of their respective owners and are used for identification only.