Référence Terraform pour ClickPipes

Tous les types de ClickPipes peuvent être créés et gérés à l’aide de la ressource clickhouse_clickpipe du fournisseur Terraform ClickHouse. Cette page présente la configuration du fournisseur ainsi que des exemples de configuration pour chaque type de ClickPipe pris en charge.

Configuration du fournisseur

Ajoutez le fournisseur ClickHouse à votre configuration Terraform :

terraform {
  required_providers {
    clickhouse = {
      source  = "ClickHouse/clickhouse"
      version = ">= 3.14.0"
    }
  }
}

provider "clickhouse" {
  organization_id = var.organization_id
  token_key       = var.token_key
  token_secret    = var.token_secret
}

Consultez Gestion des clés API pour savoir comment créer une clé API à utiliser avec le fournisseur.

Aperçu de la ressource

La ressource clickhouse_clickpipe comporte les arguments de niveau supérieur suivants :

Argument	Obligatoire	Description
`name`	Oui	Nom du ClickPipe.
`service_id`	Oui	Identifiant du service ClickHouse Cloud.
`source`	Oui	Configuration de la source (un bloc source par ClickPipe).
`destination`	Oui	Configuration de destination.
`scaling`	Non	Nombre et taille des répliques. Par défaut : 1 réplique.
`field_mappings`	Non	Mappages de champs personnalisés entre les colonnes source et de destination.
`settings`	Non	Paramètres avancés de ClickPipe.
`stopped`	Non	Définissez `true` pour créer le ClickPipe à l’état arrêté. La valeur par défaut est `false`.

Les attributs id et state sont en lecture seule et sont renseignés par ClickHouse Cloud après la création.

Destination

Le bloc destination est commun à tous les types de source :

destination {
  database      = "default"           # Target database. Defaults to "default".
  table         = "my_table"          # Target table name. Required for all sources except CDC.
  managed_table = true                # Let ClickPipes create and manage the table. Defaults to true.

  table_definition {
    engine {
      type      = "MergeTree"         # MergeTree, ReplacingMergeTree, SummingMergeTree, or Null.
    }
    sorting_key   = ["id", "ts"]      # Optional.
    partition_by  = "toYYYYMM(ts)"    # Optional.
  }

  columns {
    name = "id"
    type = "UInt64"
  }

  columns {
    name = "message"
    type = "String"
  }
}

Pour les sources CDC (Postgres, MySQL, MongoDB, BigQuery), les tables de destination sont créées automatiquement à partir du schéma de la source — seul database est généralement nécessaire.

Exemples par type de ClickPipe

Kafka

Valeurs de type prises en charge : kafka, confluent, msk, azureeventhub, redpanda, warpstream.

resource "clickhouse_clickpipe" "kafka_clickpipe" {
  name       = "My Kafka ClickPipe"
  service_id = var.service_id

  scaling {
    replicas               = 2
    replica_cpu_millicores = 250
    replica_memory_gb      = 1.0
  }

  source {
    kafka {
      type           = "confluent"
      format         = "JSONEachRow"
      brokers        = "broker.example.com:9092"
      topics         = "my_topic"
      consumer_group = "clickpipes-consumer-group"
      authentication = "PLAIN"

      credentials {
        username = "my_user"
        password = var.kafka_password
      }

      offset {
        strategy = "from_latest"
      }
    }
  }

  destination {
    table         = "my_table"
    managed_table = true

    table_definition {
      engine {
        type = "MergeTree"
      }
    }

    columns {
      name = "id"
      type = "UInt64"
    }

    columns {
      name = "message"
      type = "String"
    }

    columns {
      name = "timestamp"
      type = "DateTime"
    }
  }
}

Kafka avec un registre de schémas

resource "clickhouse_clickpipe" "kafka_avro_clickpipe" {
  name       = "My Kafka Avro ClickPipe"
  service_id = var.service_id

  source {
    kafka {
      type   = "confluent"
      format = "AvroConfluent"
      brokers = "broker.example.com:9092"
      topics  = "my_avro_topic"

      credentials {
        username = "my_user"
        password = var.kafka_password
      }

      schema_registry {
        url            = "https://schema-registry.example.com"
        authentication = "PLAIN"

        credentials {
          username = "sr_user"
          password = var.schema_registry_password
        }
      }
    }
  }

  destination {
    table         = "my_avro_table"
    managed_table = true

    table_definition {
      engine {
        type = "MergeTree"
      }
    }

    columns {
      name = "id"
      type = "UInt64"
    }

    columns {
      name = "event"
      type = "String"
    }
  }
}

Amazon Kinesis

resource "clickhouse_clickpipe" "kinesis_clickpipe" {
  name       = "My Kinesis ClickPipe"
  service_id = var.service_id

  source {
    kinesis {
      format         = "JSONEachRow"
      stream_name    = "my-stream"
      region         = "us-east-1"
      iterator_type  = "TRIM_HORIZON"
      authentication = "IAM_USER"

      access_key {
        access_key_id = var.aws_access_key_id
        secret_key    = var.aws_secret_key
      }
    }
  }

  destination {
    table         = "my_table"
    managed_table = true

    table_definition {
      engine {
        type = "MergeTree"
      }
    }

    columns {
      name = "id"
      type = "UInt64"
    }

    columns {
      name = "message"
      type = "String"
    }
  }
}

Kinesis avec un rôle IAM

L'authentification via un rôle IAM nécessite un service ClickHouse exécuté sur AWS.

resource "clickhouse_clickpipe" "kinesis_iam_role_clickpipe" {
  name       = "My Kinesis ClickPipe (IAM Role)"
  service_id = var.service_id

  source {
    kinesis {
      format         = "JSONEachRow"
      stream_name    = "my-stream"
      region         = "us-east-1"
      iterator_type  = "LATEST"
      authentication = "IAM_ROLE"
      iam_role       = "arn:aws:iam::123456789012:role/my-kinesis-role"
    }
  }

  destination {
    table         = "my_table"
    managed_table = true

    table_definition {
      engine {
        type = "MergeTree"
      }
    }

    columns {
      name = "id"
      type = "UInt64"
    }

    columns {
      name = "message"
      type = "String"
    }
  }
}

Amazon S3

resource "clickhouse_clickpipe" "s3_clickpipe" {
  name       = "My S3 ClickPipe"
  service_id = var.service_id

  source {
    object_storage {
      type           = "s3"
      url            = "https://my-bucket.s3.amazonaws.com/data/*.json"
      format         = "JSONEachRow"
      authentication = "IAM_USER"

      access_key {
        access_key_id = var.aws_access_key_id
        secret_key    = var.aws_secret_key
      }
    }
  }

  destination {
    table         = "my_table"
    managed_table = true

    table_definition {
      engine {
        type = "MergeTree"
      }
    }

    columns {
      name = "id"
      type = "UInt64"
    }

    columns {
      name = "message"
      type = "String"
    }
  }
}

Ingestion continue S3 avec SQS

Pour une ingestion continue à l’aide d’une file d’attente SQS (mode unordered). Voir Configuring unordered mode pour les instructions de configuration.

resource "clickhouse_clickpipe" "s3_continuous_clickpipe" {
  name       = "My S3 Continuous ClickPipe"
  service_id = var.service_id

  source {
    object_storage {
      type           = "s3"
      url            = "https://my-bucket.s3.amazonaws.com/data/*.json"
      format         = "JSONEachRow"
      is_continuous  = true
      queue_url      = "https://sqs.us-east-1.amazonaws.com/123456789012/my-queue"
      authentication = "IAM_USER"

      access_key {
        access_key_id = var.aws_access_key_id
        secret_key    = var.aws_secret_key
      }
    }
  }

  destination {
    table         = "my_table"
    managed_table = true

    table_definition {
      engine {
        type = "MergeTree"
      }
    }

    columns {
      name = "id"
      type = "UInt64"
    }

    columns {
      name = "message"
      type = "String"
    }
  }
}

Google Cloud Storage

service_account_key doit contenir le contenu encodé en base64 d’un fichier de clé JSON de compte de service GCP.

resource "clickhouse_clickpipe" "gcs_clickpipe" {
  name       = "My GCS ClickPipe"
  service_id = var.service_id

  source {
    object_storage {
      type                = "gcs"
      url                 = "gs://my-bucket/data/*.json"
      format              = "JSONEachRow"
      authentication      = "SERVICE_ACCOUNT"
      service_account_key = var.gcs_service_account_key
    }
  }

  destination {
    table         = "my_table"
    managed_table = true

    table_definition {
      engine {
        type = "MergeTree"
      }
    }

    columns {
      name = "id"
      type = "UInt64"
    }

    columns {
      name = "message"
      type = "String"
    }
  }
}

Azure Blob Storage

resource "clickhouse_clickpipe" "abs_clickpipe" {
  name       = "My Azure Blob ClickPipe"
  service_id = var.service_id

  source {
    object_storage {
      type                 = "azureblobstorage"
      azure_container_name = "my-container"
      path                 = "data/*.json"
      format               = "JSONEachRow"
      authentication       = "CONNECTION_STRING"
      connection_string    = var.azure_connection_string
    }
  }

  destination {
    table         = "my_table"
    managed_table = true

    table_definition {
      engine {
        type = "MergeTree"
      }
    }

    columns {
      name = "id"
      type = "UInt64"
    }

    columns {
      name = "message"
      type = "String"
    }
  }
}

Postgres CDC

resource "clickhouse_clickpipe" "postgres_cdc_clickpipe" {
  name       = "My Postgres CDC ClickPipe"
  service_id = var.service_id

  source {
    postgres {
      host     = "postgres.example.com"
      port     = 5432
      database = "mydb"

      credentials {
        username = "postgres_user"
        password = var.postgres_password
      }

      settings {
        replication_mode = "cdc"

        # Optional settings
        sync_interval_seconds              = 60
        pull_batch_size                    = 100000
        allow_nullable_columns             = true
        initial_load_parallelism           = 4
        snapshot_num_rows_per_partition    = 100000
        snapshot_number_of_parallel_tables = 1
      }

      table_mappings {
        source_schema_name = "public"
        source_table       = "users"
        target_table       = "public_users"
      }

      table_mappings {
        source_schema_name = "public"
        source_table       = "orders"
        target_table       = "public_orders"

        # Optional
        excluded_columns       = ["internal_notes"]
        use_custom_sorting_key = true
        sorting_keys           = ["id", "created_at"]
        table_engine           = "ReplacingMergeTree"
      }
    }
  }

  destination {
    database = "default"
  }
}

Postgres avec un rôle IAM

L'authentification par rôle IAM nécessite un service ClickHouse hébergé sur AWS.

resource "clickhouse_clickpipe" "postgres_iam_role_clickpipe" {
  name       = "My Postgres CDC ClickPipe (IAM Role)"
  service_id = var.service_id

  source {
    postgres {
      host           = "mydb.cluster.us-east-1.rds.amazonaws.com"
      port           = 5432
      database       = "mydb"
      type           = "rdspostgres"
      authentication = "iam_role"
      iam_role       = "arn:aws:iam::123456789012:role/my-rds-role"

      credentials {
        username = "postgres_user"
      }

      settings {
        replication_mode = "cdc"
      }

      table_mappings {
        source_schema_name = "public"
        source_table       = "orders"
        target_table       = "public_orders"
      }
    }
  }

  destination {
    database = "default"
  }
}

MySQL CDC

resource "clickhouse_clickpipe" "mysql_cdc_clickpipe" {
  name       = "My MySQL CDC ClickPipe"
  service_id = var.service_id

  source {
    mysql {
      host = "mysql.example.com"
      port = 3306
      type = "mysql"

      credentials {
        username = "mysql_user"
        password = var.mysql_password
      }

      settings {
        replication_mode = "cdc"

        # Optional settings
        sync_interval_seconds              = 30
        pull_batch_size                    = 10000
        allow_nullable_columns             = true
        initial_load_parallelism           = 4
        snapshot_num_rows_per_partition    = 100000
        snapshot_number_of_parallel_tables = 2
      }

      table_mappings {
        source_schema_name = "mydb"
        source_table       = "orders"
        target_table       = "mydb_orders"
      }

      table_mappings {
        source_schema_name = "mydb"
        source_table       = "customers"
        target_table       = "mydb_customers"

        # Optional
        excluded_columns       = ["password_hash"]
        use_custom_sorting_key = true
        sorting_keys           = ["id"]
        table_engine           = "ReplacingMergeTree"
      }
    }
  }

  destination {
    database = "default"
  }
}

MongoDB CDC

resource "clickhouse_clickpipe" "mongodb_cdc_clickpipe" {
  name       = "My MongoDB CDC ClickPipe"
  service_id = var.service_id

  source {
    mongodb {
      uri             = "mongodb+srv://cluster0.example.mongodb.net"
      read_preference = "secondaryPreferred"

      credentials {
        username = "mongo_user"
        password = var.mongodb_password
      }

      settings {
        replication_mode = "cdc"

        # Optional settings
        sync_interval_seconds              = 30
        pull_batch_size                    = 500
        snapshot_num_rows_per_partition    = 100000
        snapshot_number_of_parallel_tables = 2
      }

      table_mappings {
        source_database_name = "mydb"
        source_collection    = "users"
        target_table         = "mydb_users"
      }

      table_mappings {
        source_database_name = "mydb"
        source_collection    = "orders"
        target_table         = "mydb_orders"
        table_engine         = "ReplacingMergeTree"
      }
    }
  }

  destination {
    database = "default"
  }
}

BigQuery

service_account_file doit correspondre au contenu, encodé en base64, d’un fichier de clé JSON de compte de service GCP.

resource "clickhouse_clickpipe" "bigquery_snapshot_clickpipe" {
  name       = "My BigQuery ClickPipe"
  service_id = var.service_id

  source {
    bigquery {
      snapshot_staging_path = "gs://my-staging-bucket/staging/"

      credentials {
        service_account_file = var.gcp_service_account_key
      }

      settings {
        replication_mode = "snapshot"

        # Optional settings
        initial_load_parallelism           = 4
        snapshot_num_rows_per_partition    = 100000
        snapshot_number_of_parallel_tables = 2
        allow_nullable_columns             = true
      }

      table_mappings {
        source_dataset_name = "my_dataset"
        source_table        = "my_table"
        target_table        = "my_bigquery_table"
      }

      table_mappings {
        source_dataset_name    = "my_dataset"
        source_table           = "another_table"
        target_table           = "another_bigquery_table"
        table_engine           = "ReplacingMergeTree"
        use_custom_sorting_key = true
        sorting_keys           = ["id"]
        excluded_columns       = ["internal_col"]
      }
    }
  }

  destination {
    database = "default"
  }
}

Mise à l’échelle

Tous les types de ClickPipe prennent en charge le bloc scaling pour configurer le nombre de répliques et l’allocation des ressources :

scaling {
  replicas               = 2     # Default: 1. Maximum: 10.
  replica_cpu_millicores = 500   # Between 125 and 2000.
  replica_memory_gb      = 2.0   # Between 0.5 and 8.0.
}

Importer des ClickPipes existants

Les ClickPipes existants peuvent être importés dans le Terraform state à l’aide de l’ID du service et de l’ID du ClickPipe :

terraform import clickhouse_clickpipe.example <service_id>:<clickpipe_id>