Read and pull data from S3 object storages and web resources over HTTP
With federated queries in Aiven for ClickHouse®, you can read and pull data from an external S3-compatible object storage or any web resource accessible over HTTP. Learn more about capabilities and applications of federated queries in About querying external data in Aiven for ClickHouse®.
This article shows how to run federated queries in Aiven for ClickHouse. It provides multiple examples of querying external resources using the SELECT and INSERT statements and the S3 and the URL functions.
About running federated queries
Federated queries are written using specific SQL statements and can be run from CLI, for instance. To run a federated query, just send a query over an external S3-compatible object storage including relevant S3 bucket details. A properly-constructed federated query returns a specific output.
Before you start
Access and permissions
To run a federated query, the ClickHouse service user connecting to the cluster requires grants to the S3 and/or URL sources. The main service user is granted access to the sources by default, and new users can be allowed to use the sources with the following query:
GRANT CREATE TEMPORARY TABLE, S3, URL ON *.* TO <username> [WITH GRANT OPTION]
The CREATE TEMPORARY TABLE grant is required for both sources. Adding WITH GRANT OPTION allows the user to further transfer the privileges.
Limitations
- Federated queries in Aiven for ClickHouse only support S3-compatible object storage providers for the time being. More external data sources coming soon!
- Virtual tables are only supported for URL sources, using the URL table engine. Stay tuned for us supporting the S3 table engine in the future!
Run a federated query
Check out some examples of running federated queries to read and pull data from external S3-compatible object storages.
Query using SELECT and the S3 function
SQL SELECT statements using the S3 and URL functions are able to query public resources using the URL of the resource. For instance, let's explore the network connectivity measurement data provided by the Open Observatory of Network Interference (OONI).
WITH ooni_data_sample AS
(
SELECT *
FROM s3('https://ooni-data-eu-fra.s3.eu-central-1.amazonaws.com/clickhouse_export/csv/fastpath_202308.csv.zstd')
LIMIT 100000
)
SELECT
probe_cc AS probe_country_code,
test_name,
countIf(anomaly = 't') AS total_anomalies
FROM ooni_data_sample
GROUP BY
probe_country_code,
test_name
HAVING total_anomalies > 10
ORDER BY total_anomalies DESC
LIMIT 50
Query using SELECT and the s3Cluster function
The s3Cluster function allows all cluster nodes to participate in the
query execution. Using default for the cluster name parameter, we can
compute the same aggregations as above as follows:
WITH ooni_clustered_data_sample AS
(
SELECT *
FROM s3Cluster('default', 'https://ooni-data-eu-fra.s3.eu-central-1.amazonaws.com/clickhouse_export/csv/fastpath_202308.csv.zstd')
LIMIT 100000
)
SELECT
probe_cc AS probe_country_code,
test_name,
countIf(anomaly = 't') AS total_anomalies
FROM ooni_clustered_data_sample
GROUP BY
probe_country_code,
test_name
HAVING total_anomalies > 10
ORDER BY total_anomalies DESC
LIMIT 50
Query a private S3 bucket
Private buckets can be accessed by providing the access token and secret as function parameters.
SELECT *
FROM s3(
'https://private-bucket.s3.eu-west-3.amazonaws.com/dataset-prefix/partition-name.csv',
'some_aws_access_key_id',
'some_aws_secret_access_key'
)
Depending on the format, the schema can be automatically detected. If it isn't, you may also provide the column types as function parameters.
SELECT *
FROM s3(
'https://private-bucket.s3.eu-west-3.amazonaws.com/orders-dataset/partition-name.csv',
'access_token',
'secret_token',
'CSVWithNames',
"`order_id` UInt64, `quantity` Decimal(9, 18), `order_datetime` DateTime"
)
Query using SELECT and the URL function
Let's query the Growth Projections and Complexity Rankings dataset, courtesy of the Atlas of Economic Complexity project.
WITH economic_complexity_ranking AS
(
SELECT *
FROM url('https://dataverse.harvard.edu/api/access/datafile/7259657?format=tab', 'TSV')
)
SELECT
replace(code, '"', '') AS `ISO country code`,
growth_proj AS `Forecasted annualized rate of growth`,
toInt32(replace(sitc_eci_rank, '"', '')) AS `Economic Complexity Index ranking`
FROM economic_complexity_ranking
WHERE year = 2021
ORDER BY `Economic Complexity Index ranking` ASC
LIMIT 20
Query using INSERT and the URL function
With the URL function, INSERT statements generate a POST request, which
can be used to interact with APIs having public endpoints. For instance,
if your application has a ingest-csv endpoint accepting CSV data, you
can insert a row using the following statement:
INSERT INTO FUNCTION
url('https://app-name.company-name.cloud/api/ingest-csv', 'CSVWithNames')
VALUES ('column1-value', 'column2-value');
Query using INSERT and the S3 function
When executing an INSERT statement into the S3 function, the rows are appended to the corresponding object if the table structure matches:
INSERT INTO FUNCTION
s3('https://bucket-name.s3.region-name.amazonaws.com/dataset-name/landing/raw-data.csv', 'CSVWithNames')
VALUES ('column1-value', 'column2-value');
Query a virtual table
Instead of specifying the URL of the resource in every query, it's possible to create a virtual table using the URL table engine. This can be achieved by running a DDL CREATE statement similar to the following:
CREATE TABLE trips_export_endpoint_table
(
`trip_id` UInt32,
`vendor_id` UInt32,
`pickup_datetime` DateTime,
`dropoff_datetime` DateTime,
`trip_distance` Float64,
`fare_amount` Float32
)
ENGINE = URL('https://app-name.company-name.cloud/api/trip-csv-export', CSV)
Once the table is defined, SELECT and INSERT statements execute GET and POST requests to the URL respectively:
SELECT
toDate(pickup_datetime) AS pickup_date,
median(fare_amount) AS median_fare_amount,
max(fare_amount) AS max_fare_amount
FROM trips_export_endpoint_table
GROUP BY pickup_date
INSERT INTO trips_export_endpoint_table
VALUES (8765, 10, now() - INTERVAL 15 MINUTE, now(), 50, 20)