Event table columns

An event table is a special kind of database table with a predefined set of columns. The table’s structure is designed to support the data model for OpenTelemetry, a framework for handling telemetry data.

For more information about working with event tables, see Working with event tables.

Event table columns

Event tables have the following columns:

Column

Data Type

Description

TIMESTAMP

TIMESTAMP_NTZ

The UTC timestamp when an event was created. For events representing a span of time, this is the end of the time span.

START_TIMESTAMP

TIMESTAMP_NTZ

For events representing a span of time, such as trace events, the start of the time span as a UTC timestamp.

OBSERVED_TIMESTAMP

TIMESTAMP_NTZ

A UTC time used for logs. Currently the same value as for TIMESTAMP.

TRACE

OBJECT

Tracing context for all signal types. Contains string values trace_id and span_id.

RESOURCE

OBJECT

Reserved for future use.

RESOURCE_ATTRIBUTES

OBJECT

Attributes that identify the source of an event such as database, schema, user, warehouse, etc.

SCOPE

OBJECT

Scopes for events. For example, class names for logs.

SCOPE_ATTRIBUTES

OBJECT

Reserved for future use.

RECORD_TYPE

STRING

The event type. One of the following:

  • LOG for a log message.

  • SPAN for user-defined function invocations performed sequentially on the same thread. For more information, see RECORD_TYPE column.

  • SPAN_EVENT for a single trace event. A single query can emit more than one SPAN_EVENT.

RECORD

OBJECT

Fixed values for each record type, as described in RECORD column.

RECORD_ATTRIBUTES

OBJECT

Variable attributes for each record type, as described in RECORD_ATTRIBUTES column.

VALUE

VARIANT

Primary event value.

EXEMPLARS

ARRAY

Reserved for future use.

Data captured by event type

Data for logs

Attribute

Description

OBSERVED_TIMESTAMP

Currently the same value as for TIMESTAMP.

RECORD

The severity level recorded by the log event.

RECORD_ATTRIBUTES

The location in code from which the log event was emitted. The values vary by language, but can include the code file path, function name, line number, and so on.

RECORD_TYPE

The event type: LOG for a log message

RESOURCE_ATTRIBUTES

Attributes that identify the source of the event, such as database, schema, user, warehouse, and so on.

SCOPE

Scope within which the event occurred, such as the name of the class where the log event was created.

TIMESTAMP

The timestamp when the event was created.

VALUE

The log message.

Data for metrics

Note

Metrics data collection is a preview feature available to all accounts.

Attribute

Description

RECORD

For a metric event, an object that includes the metric’s name and unit.

RECORD_TYPE

The event type: METRIC for a metric data point.

RESOURCE_ATTRIBUTES

Attributes that identify the source of the event, such as database, schema, user, or warehouse.

START_TIMESTAMP

When the RECORD column metric_type value is sum, this is the time when the metric was collected. Not used when the metric_type value is gauge.

TIMESTAMP

The timestamp when the event was created.

VALUE

The numeric value of the metric.

Data for trace events

Attribute

Description

RECORD

For a span, an object that includes the span’s name and kind; for a span event, the object includes the span’s name.

RECORD_ATTRIBUTES

Attribute data associated with a span or span event.

RECORD_TYPE

The event type: SPAN for a span, SPAN_EVENT for a span event.

RESOURCE_ATTRIBUTES

Attributes that identify the source of the event, such as database, schema, user, warehouse, and so on.

START_TIMESTAMP

For a span, the time when the span began. Not used for a span event.

TIMESTAMP

The timestamp when the event was created.

TRACE

Identifers trace_id and span_id for a span and the span events within it.

EXEMPLARS column

Reserved for future use.

OBSERVED_TIMESTAMP column

A log’s UTC timestamp. Not used for trace events.

RECORD column

Provides core information about the event, include the log level for a log event, or the name for trace event (span or span event record).

Attributes, if any, for the record are recorded in the RECORD_ATTRIBUTES column.

Values contained by this column will vary depending on the value of the RECORD_TYPE column (LOG, SPAN or SPAN_EVENT), as described in the following sections.

For LOG RECORD_TYPE

When the RECORD_TYPE column value is LOG, the RECORD column value contains the severity of the log message. The column value may contain the following keys:

Key

Type

Description

severity_text

STRING

The text for the log severity. One of the following:

  • TRACE

  • DEBUG

  • INFO

  • WARN

  • ERROR

  • FATAL

When the log entry is for an unhandled exception, this value is the highest-severity error level for the current language runtime. For example, for code written in Python, the value is FATAL.

Example

{
  "severity_text": "INFO"
}
Copy

For METRIC RECORD_TYPE

Note

Metrics data collection is a preview feature available to all accounts.

Metrics are CPU and memory data generated by Snowflake. You can use this data to analyze resource consumption.

The execution handler language and its environment significantly affect the meaning of the metrics data. See Emitting metrics data from handler code for more information.

Key

Type

Description

metric.name

string

The name of the metric recorded by the row. One of the following:

  • process.memory.usage: Amount of memory, in bytes, consumed during execution.

  • process.cpu.utilization: CPU use. Measured differently based on the handler language.

For more information, see Emitting metrics data from handler code.

metric.unit

string

The units of the metric; for example, bytes.

metric_type

string

The OpenTelemetry Metric Point type of the metric data; for example, sum or gauge.

value_type

string

The data type of the value in the VALUE column; for example, DOUBLE or INT.

Example

{
  "metric": {
    "name": "process.memory.usage",
    "unit": "bytes"
  },
  "metric_type": "sum",
  "value_type": "INT"
}
Copy

For SPAN RECORD_TYPE

Spans represent individual executions of functions and procedures. For stored procedures there will be a single span. For user-defined functions there may be multiple spans for a single function call, depending on how Snowflake decides to schedule execution.

All spans for a given query have the same value for the trace_id key of the TRACE column.

The duration of a span is the difference between the values in the start_timestamp and timestamp columns, indicating the time of the beginning and end of the span execution, respectively.

The ID of the span and the query trace are represented in the value in the TRACE column.

Snowflake will create one span for each execution with the keys shown below:

Key

Type

Description

dropped_attributes_count

int

The number of attributes ignored after the recorded maximum has been reached.

name

string

When the executable’s handler is written in Python, this identifies the handler for the function or procedure that emitted the data. This varies by executable type, as follows:

  • Procedure: handler function name

  • User-defined function (UDF): handler function name

  • User-defined table function (UDTF): handler class name

  • Client code: name of the client-side API that began the span.

When the executable’s handler is written in a language other than Python, this is a fixed value such as snow.auto_instrumented.

kind

string

Fixed value SPAN_KIND_INTERNAL

parent_span_id

Hex string

Identifies the span of the procedure or UDF from which the current trace passed. When this value is present, it means that the current procedure or UDF call was made by another procedure in a call chain relationship. That “parent” procedure’s span_id value is the same as this parent_span_id.

snow.process.memory.usage.max

string

Optional. When present, specifies the maximum amount of memory, in bytes, used during this span’s execution.

status

string

STATUS_CODE_ERROR when the span corresponds to an unhandled exception. Otherwise, STATUS_CODE_UNSET.

In the case of user-defined functions, Snowflake may add attributes for spans to indicate the number of rows processed and emitted by the function.

For SPAN_EVENT RECORD_TYPE

Span events are event records attached to a particular span execution, described above. You can create events to fit the needs of your application. The number of span events is limited to 128.

The value of the TRACE column will identify the span in which the event was created.

Span events have a single key, name, and can have arbitrary attributes added in the RECORD_ATTRIBUTES column.

Key

Type

Description

name

string

The name of the span event.

RECORD_ATTRIBUTES column

Describes the event with metadata set by Snowflake or by code. The value will vary depending on the type of record the row contains, as described in the following sections.

For LOG RECORD_TYPE

The location in code from which the log event was emitted, including the code file path, function name, line number, and so on.

In addition to the attributes listed below, you can add your own attributes to include in the RECORD_ATTRIBUTES value.

Attribute

Type

Description

code.filepath

int

The file containing code that generated the message.

code.function

string

The name of the function that generated the message.

code.lineno

int

The line number in code that generated the message.

code.namespace

int

The namespace of code that generated the messages.

exception.message

string

The error message from an unhandled exception.

exception.type

string

The name of the class for an unhandled exception.

exception.stacktrace

string

An unhandled exception’s stack trace formatted by a language runtime.

exception.escaped

boolean

true if this entry is from an unhandled exception.

thread.id

int

The thread on which the log event was created.

thread.name

string

The thread on which the log event was created.

Example

In the following example, all attributes have been added by Snowflake except employee.id, which was added by a custom attribute.

{
  "code.filepath": "main.scala",
  "code.function": "$anonfun$new$10",
  "code.lineno": 149,
  "code.namespace": "main.main$",
  "thread.id": 1,
  "thread.name": "main"
  "employee.id": "52307953446424"
}
Copy

For SPAN RECORD_TYPE

Attributes, if any, assigned to the span when it is recorded. Attribute names and values are set by code or by Snowflake.

The following table lists attributes that might be set by Snowflake.

Attribute

Type

Description

snow.input.rows

int

The number of input rows processed by the span of the function.

snow.output.rows

int

The number of output rows successfully processed by the span of the function.

Example

Code in the following example includes attributes set by Snowflake.

{
  "snow.input.rows": 12
  "snow.output.rows": 12
}
Copy

Example

Code in the following example includes attributes set by handler code.

{
  "MyFunctionVersion": "1.1.0"
}
Copy

For SPAN_EVENT RECORD_TYPE

Attributes, if any, assigned to the span event when it is recorded. Attribute names and values may be set by Snowflake or by user code.

Example

Code in the following example includes attributes set by handler code.

{
  "mykey1": "value1",
  "mykey2": "value2"
}
Copy

RECORD_TYPE column

Specifies the kind of record described by the event table row. This column’s value identifies which of the three types of records for which then event table may contain data.

The RECORD column contains this record’s data. The RECORD_ATTRIBUTES column contains this record’s metadata, if any.

The following table lists possible values for this column.

Column Value

Description

LOG

The row represents a log entry generated by handler code.

SPAN

The row represents a span.

For a stored procedure there will be a single span. For a user-defined function, which may be parallelized, there will be a span for each thread on which the function executes. The number of threads will vary depending on multiple factors, including the size of the Snowflake warehouse in which the function executes.

A span may contain multiple span events. For more information, see Span data recorded.

SPAN_EVENT

The row represents a span event. The may be multiple span event records attached to a particular span. Your handler code may create events to fit your needs. The number of span events is limited to 128.

METRIC [1]

The row represents an observation of a metric. Multiple observations of multiple metrics can be associated with a particular span.

RESOURCE column

Reserved for future use.

RESOURCE_ATTRIBUTES column

Describes the source of an event in terms of Snowflake objects.

Attributes making up this column’s value are set by Snowflake and cannot be changed.

Resource attributes for event source

Attribute Name

Attribute Type

Description

Example

snow.database.id

int

The internal/system-generated identifier of the database containing the executable.

12345

snow.database.name

string

The name of the database containing the executable.

MY_DATABASE

snow.executable.id

int

The internal/system-generated identifier of the executable (procedure, function, SnowService, etc.) generating the event.

12345

snow.executable.name

string

The name of the executable generating the event. For example, this might be the name of the procedure, function, or Streamlit app.

MY_UDF

snow.executable.runtime.version

string

The executable language’s runtime version. This will be a value specific to the language, as described below:

  • Java: 11 or 17

  • JavaScript: No value

  • Python: From 3.9 to 3.12

  • Scala: 2.12

  • SQL: No value

procedure

snow.executable.type

string

One of the following:

  • procedure for stored procedure

  • function for a user-defined function

  • sql for an event from a single query, such as a Snowflake Scripting block.

  • SnowparkContainers for a Snowpark Container Services service.

  • streamlit for a Streamlit app

procedure

snow.owner.id

int

The internal/system-generated identifier of the role with OWNERSHIP privilege for the executable.

1234

snow.owner.name

string

The name of the role with OWNERSHIP privilege for the executable.

UDF_OWNER_RL

snow.schema.id

int

The internal/system-generated identifier of the schema containing the executable.

12345

snow.schema.name

string

The name of the schema containing the executable.

MY_SCHEMA

telemetry.sdk.language

string

The language of the resource/SDK. Snowflake uses java, scala, python, javascript and sql.

java

Resource attributes for execution environment

Attribute

Type

Description

Examples

db.user

string

For a function or procedure, the name of the user executing the function or procedure. For a Streamlit app, the name of the user who was viewing the app for a given event.

MY_USER_NAME

snow.query.id

string

The ID of the query.

01a6aeb7-0604-c466-0000-097127d13812

snow.release.version

string

The Snowflake release running when event was generated

7.9.0

snow.session.id

int

The ID of the session running the executable.

10

snow.session.role.primary.id

int

The internal/system-generated identifier of the primary role in the session.

10

snow.session.role.primary.name

string

The name of the primary role in the session.

MY_ROLE

snow.user.id

int

The internal/system-generated identifier of the user running the query.

1234

snow.warehouse.id

int

The internal/system-generated identifier of the warehouse running the query generating the event.

12345

snow.warehouse.name

string

The name of the warehouse running the query generating the event.

MY_WAREHOUSE

Resource attributes for apps

Attribute

Type

Description

Examples

snow.application.consumer.name

string

For a Snowflake Native App, the name of the consumer’s account.

CONSUMER_NAME

snow.application.consumer.organization

string

For a Snowflake Native App, the name of the consumer’s organization.

CONSUMER_ORG_NAME

snow.application.id

string

For a Snowflake Native App, the internal/system-generated identifier of the app.

ABCZN3J3

snow.application.name

string

For a Snowflake Native App, the name of the app.

MY_INSTALLED_APP_NAME

snow.application.package.name

string

For a Snowflake Native App, the name of the application package.

MY_INSTALLED_PACKAGE_NAME

snow.listing.global_name

string

For a Snowflake Native App, the internal/system-generated identifier of the listing.

GZYZN3J3

snow.listing.name

string

For a Snowflake Native App, the name of the listing.

MY_LISTING_NAME

Resource attributes for Snowflake version

Attribute

Type

Description

Examples

service.version

string

The version of the executable, where relevant. The combination of snow.version and snow.patch joined by a dot where they exist. Standard OpenTelemetry attribute.

2.3.1

snow.patch

string

The patch level of the executable running.

1

snow.version

string

The version of the executable running.

2.3

Example

{
  "db.user": "MYUSERNAME",
  "snow.database.id": 13,
  "snow.database.name": "MY_DB",
  "snow.executable.id": 197,
  "snow.executable.name": "FUNCTION_NAME(I NUMBER):ARG_NAME(38,0)",
  "snow.executable.type": "FUNCTION",
  "snow.owner.id": 2,
  "snow.owner.name": "MY_ROLE",
  "snow.query.id": "01ab0f07-0000-15c8-0000-0129000592c2",
  "snow.schema.id": 16,
  "snow.schema.name": "PUBLIC",
  "snow.session.id": 1275605667850,
  "snow.session.role.primary.id": 2,
  "snow.session.role.primary.name": "MY_ROLE",
  "snow.user.id": 25,
  "snow.warehouse.id": 5,
  "snow.warehouse.name": "MYWH",
  "telemetry.sdk.language": "python"
}
Copy

SCOPE column

For log events, the namespace of the code that emitted the event, such as the name of the class creating a log entry. This is not used for trace events.

The following table lists attributes that may be included in this column.

Scope value

Attribute

Type

Description

Examples

name

String

Namespace of code emitting the event.

com.sample.MyClass

Example

{
  "name": "com.sample.MyClass"
}
Copy

SCOPE_ATTRIBUTES column

Reserved for future use.

START_TIMESTAMP column

The time a span started as a UTC timestamp.

RECORD_TYPE Column Value

START_TIMESTAMP Value Description

LOG

Not used.

SPAN

The time the span started.

SPAN_EVENT

Not used.

METRIC [2]

When the RECORD column metric_type value is sum, this is the time when the metric was collected. Not used when the metric_type value is gauge.

TIMESTAMP column

The time an event was emitted. The value’s meaning will vary depending on the type of record the row represents, as listed in the following table:

RECORD_TYPE Column Value

TIMESTAMP Value Description

LOG

The wall-clock time that the event was emitted.

SPAN

The time at which execution concluded.

SPAN_EVENT

The wall-clock time that the event was emitted.

TRACE column

Unique identifiers representing execution for functions and procedures.

RECORD_TYPE Column Value

TRACE Value Description

LOG

Not used.

SPAN

trace_id and span_id

SPAN_EVENT

trace_id and span_id

Trace value

The following table lists attributes that may be included in this column.

Attribute

Type

Description

Examples

span_id

Hex string

A unique identifier to the threading model. Procedures, which are single-threaded, will have a single span_id value. Functions, which may be executed by Snowflake on multiple threads (such as for multiple rows), may have multiple span_id values.

When the current span is from a procedure that called another procedure or UDF in the trace, this span_id value is the same as the RECORD column parent_span_id value of the span for the procedure or UDF it called.

b4c28078330873a2

trace_id

Hex string

A unique identifier for calls made from a query. When a stored procedure is not being called in a chain of calls, each call has its own trace_id value. Within a query, calls to all functions made from the query share the same trace_id value.

When a procedure is called by another procedure or UDF in a call chain, it has the same trace_id value as other procedures and UDFs in the chain.

This value is unique for each query and will be the same for all spans within a query. You can use it for grouping events within a single query execution.

6992e9febf0b97f45b34a62e54936adb

Example

Code in the following example shows the attributes that would be present for a span or span event.

{
  "span_id": "b4c28078330873a2",
  "trace_id": "6992e9febf0b97f45b34a62e54936adb"
}
Copy

VALUE column

  • For log events, this is usually the log message. When the event logged is for an unhandled exception, the value in this column will be simply exception.

  • For metrics, this is the numeric value of the metric. [3]

Note that the VALUE column’s type is VARIANT (not STRING) so that it can have non-string values for some languages, such as JavaScript.