With batch exports, data can be exported to a Snowflake database table.
Creating the batch export
- Subscribe to data pipelines add-on in your billing settings if you haven't already.
- Click Data pipelines in the navigation and go to the exports tab in your PostHog instance.
- Click "Create export workflow".
- Select Snowflake as the batch export destination.
- Fill in the necessary configuration details.
- Finalize the creation by clicking on "Create".
- Done! The batch export will schedule its first run on the start of the next period.
Snowflake configuration
Configuring a batch export targeting Snowflake requires the following Snowflake-specific configuration values:
- User: A Snowflake user name with permissions to insert data into the provided table and, if the table is meant to be created, permissions to create the table.
- Password: The password of the Snowflake user provided.
- Role (optional): A role to assume for the required permissions.
- Account: A Snowflake account identifier without the
snowflakecomputing.com
suffix (if any). - Database: The Snowflake database where the table provided to insert data is located (and created if not present).
- Schema: The Snowflake database schema where the table provided to insert data is located (and created if not present).
- Table name: The name of a Snowflake table where to export the data.
- Warehouse: The Snowflake warehouse used for data loading.
File staging
Batch exports use Snowflake's internal table stages to stage the files copied into your Snowflake tables. We recommend using separate tables to export data from PostHog to avoid conflicting with other workflows that use internal table stages.
Note: This differs from the old PostHog Snowflake export destination which required extra configuration and used S3 or GCS as an external stage.
Models
This section describes the models that can be exported to Snowflake.
Events model
This is the default model for Snowflake batch exports. The schema of the model as created in BigQuery is:
Field | Type | Description |
---|---|---|
uuid | STRING | The unique ID of the event within PostHog |
event | STRING | The name of the event that was sent |
properties | VARIANT | A JSON object with all the properties sent along with an event |
elements | VARIANT | This field is present for backwards compatibility but has been deprecated |
people_set | VARIANT | A JSON object with any person properties sent with the $set field |
people_set_once | VARIANT | A JSON object with any person properties sent with the $set_once field |
distinct_id | STRING | The distinct_id of the user who sent the event |
team_id | INTEGER | The team_id for the event |
ip | STRING | The IP address that was sent with the event |
site_url | STRING | This field is present for backwards compatibility but has been deprecated |
timestamp | TIMESTAMP | The timestamp associated with an event |
If the provided table doesn't exist in the provided database and schema, the first batch export run will create it. This is generally the safest option, but you may also create it yourself by running the query:
CREATE TABLE IF NOT EXISTS "{database}"."{schema}"."{table_name}" ("uuid" STRING,"event" STRING,"properties" VARIANT,"elements" VARIANT,"people_set" VARIANT,"people_set_once" VARIANT,"distinct_id" STRING,"team_id" INTEGER,"ip" STRING,"site_url" STRING,"timestamp" TIMESTAMP)COMMENT = 'PostHog events table'
Note: This is the same query used by PostHog Batch Exports if the table is not present in your database. Notice that the database, schema, and table_name parameters are surrounded by double quotes to respect the casing provided during configuration.
Persons model
The schema of the model as created in BigQuery is:
The schema of the model as created in BigQuery is:
Field | Type | Description |
---|---|---|
team_id | INTEGER | The id of the project (team) the person belongs to |
distinct_id | STRING | A distinct_id associated with the person |
person_id | STRING | The id of the person associated to this (team_id , distinct_id ) pair |
properties | VARIANT | A JSON object with all the latest properties of the person |
person_version | INTEGER | The version of the person properties associated with a (team_id , distinct_id ) pair |
person_distinct_id_version | INTEGER | The version of the person to distinct_id mapping associated with a (team_id , distinct_id ) pair |
The Snowflake table will contain one row per (team_id, distinct_id)
pair, and each pair is mapped to their corresponding person_id
and latest properties
.
If the provided table doesn't exist in the provided database and schema, the first batch export run will create it. This is generally the safest option, but you may also create it yourself by running the query:
CREATE TABLE IF NOT EXISTS "{database}"."{schema}"."{table_name}" ("team_id" INTEGER,"distinct_id" STRING,"person_id" STRING,"properties" VARIANT,"person_version" INTEGER,"person_distinct_id_version" INTEGER,)COMMENT = 'PostHog persons table'
How is the persons model kept up to date?
Exporting mutable data (like the persons model) requires executing a merge operation to apply new updates to existing rows. Executing this merge operation in Snowflake involves the following steps:
- Creating a stage table.
- Inserting new data into the stage table.
- Execute a merge operation between existing table and stage table. a. Any rows that match in the final table and for which the stage table's version is higher are updated. b. Any new rows not found in the final table are inserted.
- Drop the stage table.
FAQ
Export is showing up as NULL recods in Snowflake
This is most likely because the table's schema is different from what we are sending. To verify, run this command replacing the variables:
snow sql -q "SELECT GET_DDL('TABLE', '\"<database_name>\".\"<schema_name>\".\"<table_name>\"');" --password='<password>' --account='<account>' --username='<user>' --warehouse='<warehouse>' -x --database='"<database_name>"'
and check that the output matches this
+------------------------------------------------------------------------------------------------------------------+| GET_DDL('TABLE', || '"DATABASE_NAME"."SCHEMA_NAME"."TABLE_NAME"') ||------------------------------------------------------------------------------------------------------------------|| create or replace TABLE "table_name" ( || "uuid" VARCHAR(16777216), || "event" VARCHAR(16777216), || "properties" VARIANT, || "elements" VARIANT, || "people_set" VARIANT, || "people_set_once" VARIANT, || "distinct_id" VARCHAR(16777216), || "team_id" NUMBER(38,0), || "ip" VARCHAR(16777216), || "site_url" VARCHAR(16777216), || "timestamp" TIMESTAMP_NTZ(9) || )COMMENT='PostHog generated events table' || ; |+------------------------------------------------------------------------------------------------------------------+
If your table doesn't match exactly (including casing), then the recommended approach is delete the table and let PostHog create it for you (alternatively you can copy the DDL exactly).
Export shows a "no active warehouse selected" error
This error indicates the warehouse provided is not active (likely SUSPENDED
). You will have to inspect your Snowflake configuration to discover why. One potential explanation is that Snowflake's auto-suspend feature suspended the warehouse after some time of inactivity, which could be the case if the warehouse is used only for batch exports on slow frequencies (like daily).
Batch exports require a RUNNING
Snowflake warehouse to be able to export events to. It does not activate, suspend, or manipulate warehouses in any capacity, which can be enforced by not giving permissions required to do so to the user provided to batch exports. The reason for this is that a running warehouse consumes Snowflake credits, and deciding on Snowflake credit spending is out of scope for batch exports.
Once the cause has been identified and the configuration amended, a warehouse can be started with the query ALTER WAREHOUSE <name> RESUME
or via the Snowflake console.
See Snowflake's documentation on managing warehouses and the cost management auto-suspend feature for more details.
Examples
These examples illustrate how to use the data from batch exports in Snowflake.
Requirements
Two batch exports need to be created:
- An events model batch export.
- A persons model batch export.
For the purposes of these examples, assume that these two batch exports have already been created and have exported some data to Snowflake in tables example.events
and example.persons
.
Example: Count unique persons that have triggered an event
The following query can be used to count the number of unique persons that have triggered events:
SELECTevent,COUNT(DISTINCT persons.person_id) AS unique_persons_countFROMexample.events AS eventsLEFT JOINexample.persons AS persons ON events.distinct_id = persons.distinct_id AND events.team_id = persons.team_idWHEREpersons.person_id IS NOT NULLGROUP BYeventORDER BYunique_persons_count DESC