Contact UsGet started

Managing machine learning models in Managed Service for ClickHouse®

Written by
Updated at October 17, 2024

Managed Service for ClickHouse® allows you to analyze data by applying CatBoost machine learning models without additional tools.

To apply a model, add it to your cluster and call it in an SQL query using the built-in catboostEvaluate() function. After running this query, you get model predictions for each row of input data.

Read more about the catboostEvaluate() function in the ClickHouse® documentation.

Before adding a model

Managed Service for ClickHouse® only works with readable models uploaded to Yandex Object Storage:

  1. Upload the trained model file to Yandex Object Storage.

  2. Configure access to the model file using a service account:

    1. Connect a service account to a cluster.
    2. Assign the service account the storage.viewer role.
    3. In the bucket ACL, grant the READ permission to the account.

  3. Get a link to the model file.

Getting a list of models in a cluster

  1. In the management console, go to the folder page and select Managed Service for ClickHouse.
  2. Click the cluster name and select the Machine learning tab in the left-hand panel.

If you do not have the Yandex Cloud command line interface yet, install and initialize it.

The folder specified in the CLI profile is used by default. You can specify a different folder using the --folder-name or --folder-id parameter.

To get a list of models in a cluster, run the command:

yc managed-clickhouse ml-model list --cluster-name=<cluster_name>

You can request the cluster name with a list of clusters in the folder.

To get a list of models in a cluster, use the list REST API method for the MlModel resource or the MlModelService/List gRPC API call and provide the cluster ID in the clusterId request parameter.

You can get the cluster ID with a list of clusters in the folder.

Getting detailed information about a model

  1. In the management console, go to the folder page and select Managed Service for ClickHouse.
  2. Click the cluster name and select the Machine learning tab in the left-hand panel.

If you do not have the Yandex Cloud command line interface yet, install and initialize it.

The folder specified in the CLI profile is used by default. You can specify a different folder using the --folder-name or --folder-id parameter.

To get model details, run this command:

yc managed-clickhouse ml-model get <model_name> \
  --cluster-name=<cluster_name>

You can request the model name with a list of cluster models and the cluster name with a list of clusters in the folder.

To get model details, use the get REST API method for the MlModel resource or the MlModelService/Get gRPC API call and provide the following in the request:

  • Cluster ID in the clusterId parameter.
  • Model name in the mlModelName parameter.

You can request the model name with a list of cluster models and the cluster name with a list of clusters in the folder.

Creating a model

Note

The only supported model type is CatBoost: ML_MODEL_TYPE_CATBOOST.

  1. Select the cluster:

    1. In the management console, go to the folder page and select Managed Service for ClickHouse.
    2. Click the cluster name and select the Machine learning tab in the left-hand panel.
    3. Click Create model.

  2. Configure the model parameters:

    • Type: ML_MODEL_TYPE_CATBOOST.
    • Name: Model name. Model name is one of the arguments of the catboostEvaluate() function, which is used to call the model in ClickHouse®.
    • URL: Model address in Yandex Object Storage.

  3. Click Create and wait for the model to be created.

If you do not have the Yandex Cloud command line interface yet, install and initialize it.

The folder specified in the CLI profile is used by default. You can specify a different folder using the --folder-name or --folder-id parameter.

To create a model, run this command:

yc managed-clickhouse ml-model create <model_name> \
  --cluster-name=<cluster_name> \
  --type=ML_MODEL_TYPE_CATBOOST \
  --uri=<link_to_model_file_in_Object_Storage>

You can request the cluster name with a list of clusters in the folder.

  1. Open the current Terraform configuration file with an infrastructure plan.

    For more information about creating this file, see Creating clusters.

  2. To the Managed Service for ClickHouse® cluster description, add the ml_model block with a description of the added machine learning model:

    resource "yandex_mdb_clickhouse_cluster" "<cluster_name>" {
  ...
  ml_model {
    name = "<model_name>"
    type = "ML_MODEL_TYPE_CATBOOST"
    uri  = "<link_to_model_file_in_Object_Storage>"
  }
}

  3. Make sure the settings are correct.

    1. Using the command line, navigate to the folder that contains the up-to-date Terraform configuration files with an infrastructure plan.

    2. Run the command:

      terraform validate

      If there are errors in the configuration files, Terraform will point to them.

  4. Confirm updating the resources.

    1. Run the command to view planned changes:

      terraform plan

      If the resource configuration descriptions are correct, the terminal will display a list of the resources to modify and their parameters. This is a test step. No resources are updated.

    2. If you are happy with the planned changes, apply them:

      1. Run the command:

        terraform apply

      2. Confirm the update of resources.

      3. Wait for the operation to complete.

For more information, see the Terraform provider documentation.

Time limits

A Terraform provider sets the timeout for Managed Service for ClickHouse® cluster operations:

  • Creating a cluster, including by restoring one from a backup: 60 minutes.
  • Editing a cluster: 90 minutes.
  • Deleting a cluster: 30 minutes.

Operations exceeding the set timeout are interrupted.

How do I change these limits?

Add the timeouts block to the cluster description, for example:

resource "yandex_mdb_clickhouse_cluster" "<cluster_name>" {
  ...
  timeouts {
    create = "1h30m" # 1 hour 30 minutes
    update = "2h"    # 2 hours
    delete = "30m"   # 30 minutes
  }
}

To create a model, use the create REST API method for the MlModel resource or the MlModelService/Create gRPC API call and provide the following in the request:

  • Cluster ID in the clusterId parameter.
  • Model name in the mlModelName parameter.
  • ML_MODEL_TYPE_CATBOOST model type in the type parameter.
  • Link to the model file in Yandex Object Storage in the uri parameter.

You can get the cluster ID with a list of clusters in the folder.

Applying a model

To apply the model to data stored in a ClickHouse® cluster:

  1. Connect to the cluster.

  2. Execute an SQL query in the format:

    SELECT
    catboostEvaluate('<model_file_path>',
                  <column_1_name>,
                  <column_2_name>,
                  ...
                  <column_N_name>)
FROM <table_name>

As the catboostEvaluate() function arguments, specify the following:

  • Path to the model file formatted as /var/lib/clickhouse/models/<model_name>.bin.
  • Names of columns containing the input data.

The result of the query execution will be a column with model predictions for each row of the source table.

Updating a model

Managed Service for ClickHouse® does not track changes in the model file located in the Yandex Object Storage bucket.

To update the contents of a model that is already connected to the cluster:

  1. Upload the file with the current model to Yandex Object Storage.
  2. Get a link to this file.
  3. Change the parameters of the model connected to Managed Service for ClickHouse® by providing a new link to the model file.
  1. In the management console, go to the folder page and select Managed Service for ClickHouse.
  2. Click the cluster name and select the Machine learning tab in the left-hand panel.
  3. Select the model you need. Then click and select Edit.

If you do not have the Yandex Cloud command line interface yet, install and initialize it.

The folder specified in the CLI profile is used by default. You can specify a different folder using the --folder-name or --folder-id parameter.

To change the link to the model file in the Yandex Object Storage bucket, run the command:

yc managed-clickhouse ml-model update <model_name> \
  --cluster-name=<cluster_name> \
  --uri=<new_link_to_file_in_Object_Storage>

You can request the model name with a list of cluster models and the cluster name with a list of clusters in the folder.

  1. Open the current Terraform configuration file with an infrastructure plan.

    For more information about creating this file, see Creating clusters.

  2. In the Managed Service for ClickHouse® cluster description, change the uri parameter value under ml_model:

    resource "yandex_mdb_clickhouse_cluster" "<cluster_name>" {
...
  ml_model {
    name = "<model_name>"
    type = "ML_MODEL_TYPE_CATBOOST"
    uri  = "<new_link_to_model_file_in_Object_Storage>"
  }
}

  3. Make sure the settings are correct.

    1. Using the command line, navigate to the folder that contains the up-to-date Terraform configuration files with an infrastructure plan.

    2. Run the command:

      terraform validate

      If there are errors in the configuration files, Terraform will point to them.

  4. Confirm updating the resources.

    1. Run the command to view planned changes:

      terraform plan

      If the resource configuration descriptions are correct, the terminal will display a list of the resources to modify and their parameters. This is a test step. No resources are updated.

    2. If you are happy with the planned changes, apply them:

      1. Run the command:

        terraform apply

      2. Confirm the update of resources.

      3. Wait for the operation to complete.

For more information, see the Terraform provider documentation.

Time limits

A Terraform provider sets the timeout for Managed Service for ClickHouse® cluster operations:

  • Creating a cluster, including by restoring one from a backup: 60 minutes.
  • Editing a cluster: 90 minutes.
  • Deleting a cluster: 30 minutes.

Operations exceeding the set timeout are interrupted.

How do I change these limits?

Add the timeouts block to the cluster description, for example:

resource "yandex_mdb_clickhouse_cluster" "<cluster_name>" {
  ...
  timeouts {
    create = "1h30m" # 1 hour 30 minutes
    update = "2h"    # 2 hours
    delete = "30m"   # 30 minutes
  }
}

To update a model, use the update REST API method for the MlModel resource or the MlModelService/Update gRPC API call and provide the following in the request:

  • Cluster ID in the clusterId parameter. To find out the cluster ID, get a list of clusters in the folder.
  • Model name in the mlModelName parameter.
  • New link to the model file in Yandex Object Storage in the uri parameter.
  • List of cluster configuration fields to update in the updateMask parameter.

Warning

This API method overrides all parameters of the object being modified that were not explicitly passed in the request to the default values. To avoid this, list the settings you want to change in the updateMask parameter (one line separated by commas).

Disabling a model

Note

After disabling a model, the corresponding object is kept in the Yandex Object Storage bucket. If you no longer need this model object, you can delete it.

  1. In the management console, go to the folder page and select Managed Service for ClickHouse.
  2. Click the cluster name and select the Machine learning tab in the left-hand panel.
  3. Select the model you need. Then click and select Delete.

If you do not have the Yandex Cloud command line interface yet, install and initialize it.

The folder specified in the CLI profile is used by default. You can specify a different folder using the --folder-name or --folder-id parameter.

To disable a model, run the command:

yc managed-clickhouse ml-model delete <model_name> \
  --cluster-name=<cluster_name>

You can request the model name with a list of cluster models and the cluster name with a list of clusters in the folder.

  1. Open the current Terraform configuration file with an infrastructure plan.

    For more information about creating this file, see Creating clusters.

  2. Delete the description block of the appropriate ml_model model from the Managed Service for ClickHouse® cluster description.

  3. Make sure the settings are correct.

    1. Using the command line, navigate to the folder that contains the up-to-date Terraform configuration files with an infrastructure plan.

    2. Run the command:

      terraform validate

      If there are errors in the configuration files, Terraform will point to them.

  4. Confirm updating the resources.

    1. Run the command to view planned changes:

      terraform plan

      If the resource configuration descriptions are correct, the terminal will display a list of the resources to modify and their parameters. This is a test step. No resources are updated.

    2. If you are happy with the planned changes, apply them:

      1. Run the command:

        terraform apply

      2. Confirm the update of resources.

      3. Wait for the operation to complete.

For more information, see the Terraform provider documentation.

Time limits

A Terraform provider sets the timeout for Managed Service for ClickHouse® cluster operations:

  • Creating a cluster, including by restoring one from a backup: 60 minutes.
  • Editing a cluster: 90 minutes.
  • Deleting a cluster: 30 minutes.

Operations exceeding the set timeout are interrupted.

How do I change these limits?

Add the timeouts block to the cluster description, for example:

resource "yandex_mdb_clickhouse_cluster" "<cluster_name>" {
  ...
  timeouts {
    create = "1h30m" # 1 hour 30 minutes
    update = "2h"    # 2 hours
    delete = "30m"   # 30 minutes
  }
}

To delete a model, use the delete REST API method for the MlModel resource or the MlModelService/Delete gRPC API call and provide the following in the request:

  • Cluster ID in the clusterId parameter.
  • Model name in the mlModelName parameter.

You can request the cluster ID with a list of clusters in the folder and model name with a list of models in the cluster.

Example

If you do not have a suitable data set or model to process it, you can test machine learning in Managed Service for ClickHouse® using this example. We prepared a data file for it and trained a model to analyze it. You can upload data to ClickHouse® and see model predictions for different rows of the table.

Note

In this example, we are going to use public data from the Amazon Employee Access Challenge. The model is trained to predict values in the ACTION column. The same data and model are used on GitHub.

To upload data to ClickHouse® and test the model:

  1. In the management console, add the test model:

    • Type: ML_MODEL_TYPE_CATBOOST
    • Name: ml_test
    • URL: https://storage.yandexcloud.net/managed-clickhouse/catboost_model.bin

  2. Download the file with data to analyze.

  3. Connect to the cluster.

  4. Create a test table:

    CREATE TABLE
            ml_test_table (date Date MATERIALIZED today(), 
                          ACTION UInt8, 
                          RESOURCE UInt32, 
                          MGR_ID UInt32, 
                          ROLE_ROLLUP_1 UInt32, 
                          ROLE_ROLLUP_2 UInt32, 
                          ROLE_DEPTNAME UInt32, 
                          ROLE_TITLE UInt32, 
                          ROLE_FAMILY_DESC UInt32, 
                          ROLE_FAMILY UInt32, 
                          ROLE_CODE UInt32) 
            ENGINE = MergeTree() 
PARTITION BY date 
ORDER BY date;

  5. Upload the data to the table:

    INSERT INTO ml_test_table FROM INFILE '<file_path>/train.csv' FORMAT CSVWithNames;

  6. Test the model:

    • Get predicted values in the ACTION column for the first 10 rows in the table:

      SELECT
    catboostEvaluate('/var/lib/clickhouse/models/ml_test.bin',
                    RESOURCE,
                    MGR_ID,
                    ROLE_ROLLUP_1,
                    ROLE_ROLLUP_2,
                    ROLE_DEPTNAME,
                    ROLE_TITLE,
                    ROLE_FAMILY_DESC,
                    ROLE_FAMILY,
                    ROLE_CODE) > 0 AS prediction,
    ACTION AS target
FROM ml_test_table
LIMIT 10;

    • Get the probability prediction for the first 10 rows in the table:

      SELECT
    catboostEvaluate('/var/lib/clickhouse/models/ml_test.bin',
                    RESOURCE,
                    MGR_ID,
                    ROLE_ROLLUP_1,
                    ROLE_ROLLUP_2,
                    ROLE_DEPTNAME,
                    ROLE_TITLE,
                    ROLE_FAMILY_DESC,
                    ROLE_FAMILY,
                    ROLE_CODE) AS prediction,
    1. / (1 + exp(-prediction)) AS probability,
    ACTION AS target
FROM ml_test_table
LIMIT 10;

ClickHouse® is a registered trademark of ClickHouse, Inc.

Previous
Managing data format schemas
Next
Setting up access to Object Storage