COPY TO

Reference and syntax for the COPY TO command that exports or unloads data from Firebolt to Amazon S3.

Exports the results of a SELECT query to Amazon S3 in your specified file format. You can specify the output file format, compression type, and other options to control how the data is written, such as file name conventions and whether to overwrite existing files. Use COPY TO for backing up query results or integrating Firebolt with other data processing tools.

Syntax

COPY (<select_query>)
  TO <location_name>
  [ WITH ] ( <option> [, ...] )

COPY (<select_query>)
  TO '<url>'
  [ WITH ] ( <option> [, ...] )

<option>:
    [ CREDENTIALS = ( <credentials> ) ] 
    [ TYPE = CSV | TSV | JSON | PARQUET ]
    [ COMPRESSION = GZIP | SNAPPY | NONE ]
    [ INCLUDE_QUERY_ID_IN_FILE_NAME = TRUE | FALSE ]
    [ FILE_NAME_PREFIX = <string> ]
    [ SINGLE_FILE = TRUE | FALSE ]
    [ MAX_FILE_SIZE = <bytes> ]
    [ OVERWRITE_EXISTING_FILES = TRUE | FALSE ]
    [ HEADER = TRUE | FALSE ]
    [ <settings> ]


<credentials>: (    -- Only applicable when using URL destination
  AWS_ACCESS_KEY_ID = '<aws_access_key_id>',
  AWS_SECRET_ACCESS_KEY = '<aws_secret_access_key>'
  [, AWS_SESSION_TOKEN = '<aws_session_token>' ]
  |
  AWS_ROLE_ARN = '<aws_role_arn>'
  [, AWS_ROLE_EXTERNAL_ID = '<aws_role_external_id>' ]
)

The WITH keyword before the options list is optional. Additionally, you can use either = or a space to separate option names from their values. For example, both TYPE = JSON and TYPE JSON are valid.

Parameters

Parameter	Description
<select_query>	Any valid SELECT statement.
<location_name>	The name of a location object that contains the Amazon S3 URL and credentials. This is the recommended approach for specifying the destination. See CREATE LOCATION for more information.
<URL>	The path to an Amazon S3 or Google Cloud Storage URL where the query result file or files are saved. For example, s3://my_bucket/my_folder or gs://my_bucket/my_folder.
CREDENTIALS	The credentials for accessing the specified <URL>. For more information, see CREDENTIALS.
TYPE	Specifies the file type to save to Amazon S3. If omitted, CSV is the default.
COMPRESSION	Specifies the compression algorithm to use for the output files. If omitted, defaults to GZIP. If NONE is specified, exported files are not compressed. For TYPE=PARQUET, COMPRESSION can be set to NONE, the default GZIP, or SNAPPY. Each output file is written in Parquet format, using the specified algorithm to compress each column according to the Parquet specification. This kind of compression is internal to the Parquet file. For TYPE=CSV, TSV, or JSON, COMPRESSION can be set to NONE or the default GZIP. With GZIP, the file is first created in the specified format and then compressed into a GZIP file.
INCLUDE_QUERY_ID_IN_FILE_NAME	Specifies whether a query ID is included in the file name. Each time the statement runs, Firebolt generates a new query ID. If omitted, defaults to TRUE, and Firebolt saves file names using the pattern <query_id>.[type].gz, for example, 123ABCXY2.csv.gz. This allows Firebolt to generate unique file names by default. If FALSE, and FILE_NAME_PREFIX is not specified, files are exported with a generic output file name, for example, output.csv.gz.
FILE_NAME_PREFIX	Specifies an optional string to use in the file name. If FILE_NAME_PREFIX is omitted and INCLUDE_QUERY_ID_IN_FILE_NAME is TRUE, the exported file name is in the pattern <query_id><file_string>.<type>.gz. If FILE_NAME_PREFIX is specified and INCLUDE_QUERY_ID_IN_FILE_NAME is set to FALSE, the specified string replaces output and the file is in the pattern <file_string>.<type>.gz.
SINGLE_FILE	Specifies whether the export should be a single file or multiple files. If omitted, the default is FALSE, and the export is split based on the MAX_FILE_SIZE value. Exported files are appended with _<n> incrementally to indicate the position in series, starting with 0. For example, the first file in a series might be named 123ABCXY2_0.parquet.gz. If TRUE, only a single file is written. If set to TRUE and the file exceeds MAX_FILE_SIZE, an error occurs.
MAX_FILE_SIZE	Specifies the target file size in bytes. If omitted, the default value is 134,217,728 bytes (128 MiB). Firebolt starts a new file once the current one has passed this size, so files usually come out over it rather than under. With SINGLE_FILE = TRUE it is instead a hard limit, and the query fails if the single file exceeds it. 0 means unlimited, so Firebolt never rolls over to a new file and each writer produces a single file. Regardless of this setting, no single file written through a multipart upload can exceed 52,428,800,000 bytes (50,000 MiB), because such an upload accepts at most 10,000 parts and Firebolt uploads 5 MiB parts. That bounds every s3:// destination, and a gs:// destination that supplies an HMAC key. A gs:// destination that uses the engine identity writes each file in one resumable session, which has no part limit.
OVERWRITE_EXISTING_FILES	Specifies whether exported files should overwrite existing files of the same name in the specified Amazon S3 URL. If omitted, defaults to FALSE, and files of the same name are not overwritten.
HEADER	Specifies whether CSV files should contain a header line containing the column names. If omitted, defaults to TRUE. Must be omitted for TSV, JSON, or Parquet files.
<settings>	List of query-specific settings overrides.

For TYPE = PARQUET, Firebolt ends a row group at whichever comes first: 1,000,000 rows, or 33,554,432 bytes (32 MiB) of data. The row count is a limit that no row group exceeds; the byte figure is a target that a row group can overshoot, and it counts the data before compression, so row groups in the exported file are smaller than 32 MiB. Larger row groups let a reader skip more of a file at once, which is why row group size affects how efficiently the exported files can be queried later. Firebolt only starts a new file at a row group boundary, so a file can exceed MAX_FILE_SIZE by up to the size of the last row group it wrote, and the final file from each writer contains only the remaining rows and can be much smaller.

Credentials

Firebolt needs permissions to write query results to the specified Amazon S3 URL. You can specify IAM credentials using the AWS access keys, and the specified credentials must be associated with a user with permissions to write objects to the bucket.

Credentials for Amazon S3

Specify access key credentials using the syntax shown below.

  CREDENTIALS = ( AWS_ACCESS_KEY_ID = '<aws_access_key_id>', AWS_SECRET_ACCESS_KEY = '<aws_secret_access_key>' [, AWS_SESSION_TOKEN = '<aws_session_token>' ] | AWS_ROLE_ARN = '<aws_role_arn>' [, AWS_ROLE_EXTERNAL_ID = '<aws_role_external_id>' ] )

For Amazon S3 locations, you can use either access key-based authentication or role-based authentication:

Parameter	Description
AWS_ACCESS_KEY_ID	Your AWS access key ID
AWS_SECRET_ACCESS_KEY	Your AWS secret access key
AWS_SESSION_TOKEN	Optional temporary session token for temporary credentials
AWS_ROLE_ARN	The ARN of the IAM role to assume
AWS_ROLE_EXTERNAL_ID	Optional external ID for role assumption

For role-based AWS access you can additionally set an external ID. An external ID is a value you choose and control that AWS checks when Firebolt assumes your role, adding a second condition on top of your account’s unique IAM principal. Configuring one is a recommended best practice. See IAM roles.

For more information on how to create access keys, see Creating Access Key and Secret ID.

Credentials for Google Cloud Storage

A gs:// destination supports two credential modes:

Mode	How to specify it	How Firebolt authenticates	Principal to grant the permissions below to
HMAC key	CREDENTIALS = ( AWS_ACCESS_KEY_ID = '<hmac_access_id>', AWS_SECRET_ACCESS_KEY = '<hmac_secret>' )	The Google Cloud Storage HMAC key goes in the AWS-named fields. Firebolt writes through the S3-compatible XML API.	The service account or user the key belongs to.
Engine identity	Omit CREDENTIALS	Firebolt writes through the Google Cloud Storage gRPC API. Where the deployment sets storage.gcp.intermediary_service_account_id, the engine impersonates that service account for every external gs:// write; where it sets none, the engine authenticates as its own Google identity.	The intermediary service account where the deployment sets one, and the engine’s own Google identity where it sets none.

Whether the engine-identity mode is available depends on the deployment. Where it sets storage.gcp.intermediary_service_account_id, that service account is the principal, so the mode is available whatever storage.gcp.allow_engine_identity says: that setting governs only the engine’s own identity. Where it sets none, the engine authenticates as its own Google identity, which the deployment permits with storage.gcp.allow_engine_identity = true. The Firebolt Operator and the Helm chart enable it by default for the engines they deploy, so a self-managed engine on Kubernetes has it. A standalone engine has it once you enable it. On the Firebolt managed service the engine’s Google identity belongs to Firebolt rather than to your account, so the mode is not available there. When the mode is not available, a gs:// destination without CREDENTIALS fails with an error that asks for an HMAC key.

Both modes need storage.objects.list, because Firebolt lists the destination prefix to confirm the bucket is reachable before it exports anything, and storage.objects.create. With OVERWRITE_EXISTING_FILES = FALSE they also need storage.objects.get, because Firebolt reads each destination object’s metadata to confirm that the object does not already exist. With OVERWRITE_EXISTING_FILES = TRUE they need storage.objects.delete instead, because Google Cloud Storage counts replacing an existing object as a delete. On top of that:

• HMAC key. An object that outgrows a single 5 MiB upload part is written through the XML API’s multipart upload, which additionally needs storage.multipartUploads.create, storage.multipartUploads.abort, and storage.multipartUploads.listParts. The default MAX_FILE_SIZE is far above one part, so grant these unless every exported object is small enough for a single request.
• Engine identity. The gRPC API writes each object in one resumable session, whatever its size, so no multipart permissions apply.

MAX_FILE_SIZE applies to a gs:// destination exactly as it does to s3://, and the 50,000 MiB ceiling described in its row above binds the HMAC-key mode only.

AWS_SESSION_TOKEN and AWS_ROLE_ARN are rejected for gs:// destinations.

Least privileged permissions

The example AWS IAM policy statement below demonstrates the minimum actions that must be allowed for Firebolt to write query files to an example Amazon S3 URL. A permissions policy that allows at least these actions for the <s3_url> that you specify in the COPY TO statement must be attached to the user or role specified in the CREDENTIALS clause.

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Effect": "Allow",
            "Action": [
                "s3:Get*",
                "s3:List*",
                "s3:PutObject",
                "s3:DeleteObject"
            ],
            "Resource": [
                "arn:aws:s3:::my_s3_bucket",
                "arn:aws:s3:::my_s3_bucket/*"
            ]
        }
    ]
}

Using Location Objects

Location objects provide a secure and reusable way to specify Amazon S3 destinations for your data exports. This is the recommended approach as it:

• Centralizes credential management
• Eliminates the need to specify credentials in each COPY statement
• Provides better security through role-based access control
• Simplifies maintenance and updates

For a comprehensive guide to LOCATION objects, see LOCATION objects. For complete syntax reference, see CREATE LOCATION.

Example with Location Object

COPY (
  SELECT * FROM test_table
)
TO my_export_location
WITH (
  TYPE = JSON,
  COMPRESSION = NONE,
  SINGLE_FILE = TRUE,
  FILE_NAME_PREFIX = 'latest-export',
  MAX_FILE_SIZE = 5000000000,
  OVERWRITE_EXISTING_FILES = TRUE
);

The location object my_export_location contains the Amazon S3 URL and credentials, making the statement cleaner and more secure.

Examples

• COPY TO with location object
• COPY TO with defaults and role ARN
• COPY TO with single file and AWS access keys
• COPY TO with custom file name prefix and ARN with external ID
• COPY TO with custom file name and overwrite set to TRUE

COPY TO with location object

This example demonstrates the recommended approach using a location object. The location object contains the Amazon S3 URL and credentials, making the statement more secure and maintainable.

COPY (
  SELECT * FROM test_table
)
TO my_export_location
WITH (
  TYPE = JSON,
  COMPRESSION = NONE,
  SINGLE_FILE = TRUE,
  FILE_NAME_PREFIX = 'latest-export',
  MAX_FILE_SIZE = 5000000000,
  OVERWRITE_EXISTING_FILES = TRUE
);

Firebolt writes a single file as shown below.

s3://my_bucket/exports/
  latest-export.json

COPY TO with defaults and role ARN

The example below shows a COPY TO statement with minimal parameters that specifies an AWS_ROLE_ARN together with a recommended AWS_ROLE_EXTERNAL_ID. Because TYPE is omitted, the file or files will be written in CSV format, and because COMPRESSION is omitted, they are compressed using GZIP (*.csv.gz).

COPY (
  SELECT * FROM test_table
)
TO 's3://my-bucket/path/to/data'
WITH (
  CREDENTIALS = (AWS_ROLE_ARN='arn:aws:iam::123456789012:role/my-firebolt-role', AWS_ROLE_EXTERNAL_ID='my-external-id')
);

AWS_ROLE_EXTERNAL_ID is optional; omit it to assume the role without an external ID.

Firebolt assigns a UUID formatted query ID, to the query at runtime. The compressed output is 500 MB, exceeding the default MAX_FILE_SIZE of 128 MiB, so Firebolt writes 4 files as shown below.

s3://my_bucket/my_fb_queries/
  153831d6-c400-4014-aa58-7bdce3918019_0.csv.gz
  153831d6-c400-4014-aa58-7bdce3918019_1.csv.gz
  153831d6-c400-4014-aa58-7bdce3918019_2.csv.gz
  153831d6-c400-4014-aa58-7bdce3918019_3.csv.gz

COPY TO with single file and AWS access keys

The example below exports a single, uncompressed JSON file with the same name as the query ID. If the file to be written exceeds the specified MAX_FILE_SIZE (5 GB), an error occurs. AWS access keys are specified for credentials.

COPY (
  SELECT * FROM test_table
)
TO 's3://my-bucket/path/to/data'
WITH (
  TYPE=JSON,
  COMPRESSION=NONE,
  CREDENTIALS=(AWS_ACCESS_KEY_ID='AKIAIOSFODNN7EXAMPLE', AWS_SECRET_ACCESS_KEY='wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY'),
  INCLUDE_QUERY_ID_IN_FILE_NAME=TRUE,
  SINGLE_FILE=TRUE,
  MAX_FILE_SIZE=5000000000
);

Firebolt writes a single file as shown below.

s3://my_bucket/my_fb_queries/
  16B90E96716236D0.json

COPY TO with custom file name prefix and ARN with external ID

In the example below, because FILE_NAME_PREFIX parameter is specified, Firebolt adds a string to the query ID to form file names. The IAM role specified for CREDENTIALS also has an external ID configured in AWS, so the ID is specified.

COPY (
  SELECT * FROM test_table
)
TO 's3://my-bucket/path/to/data'
WITH (
  TYPE=JSON,
  COMPRESSION=NONE,
  CREDENTIALS = (AWS_ROLE_ARN='arn:aws:iam::123456789012:role/my-firebolt-role', AWS_ROLE_EXTERNAL_ID='99291'),
  FILE_NAME_PREFIX='_query_result'
);

Firebolt assigns the query an id of 16B90E96716236D6 at runtime and writes files as shown below.

s3://my_bucket/my_fb_queries/
  16B90E96716236D6_query_result_0.json
  16B90E96716236D6_query_result_1.json
  16B90E96716236D6_query_result_2.json
  16B90E96716236D6_query_result_3.json
  16B90E96716236D6_query_result_4.json

COPY TO with custom file name and overwrite set to TRUE

In the example below, INCLUDE_QUERY_ID_IN_FILE_NAME is set to FALSE and a FILE_NAME_PREFIX is specified. In this case, Firebolt writes files using only the specified FILE_NAME_PREFIX as the file name. In addition, SINGLE_FILE and OVERWRITE_EXISTING_FILES are set to TRUE so that the Amazon S3 URL always contains a single file with the latest query results.

COPY (
  SELECT * FROM test_table
)
TO 's3://my_bucket/my_fb_query'
WITH (
  TYPE=JSON,
  COMPRESSION=NONE,
  CREDENTIALS=(AWS_ROLE_ARN = 'arn:aws:iam::123456789012:role/my-firebolt-role'),
  INCLUDE_QUERY_ID_IN_FILE_NAME=FALSE,
  SINGLE_FILE=TRUE,
  FILE_NAME_PREFIX='latest-fb-query-result',
  MAX_FILE_SIZE=5000000000,
  OVERWRITE_EXISTING_FILES=TRUE
);

Firebolt writes a single file as shown below.

s3://my_bucket/my_fb-query/
  latest-fb-query-result.json