clickhouse-s3-etl-tools

clickhouse-s3-etl-tools

clickhouse-s3-etl-tools is a powerful utility designed for seamless data transfers between ClickHouse clusters using the flexibility of Amazon S3 or any S3-compatible storage as an intermediate staging area.

Introduction

Managing and orchestrating data movement between different ClickHouse clusters can be a challenging task. The clickhouse-s3-etl-tools service simplifies this process by acting as a bridge, enabling efficient transfers via S3 storage. This toolset is especially useful in scenarios where you need to synchronize or backup data between ClickHouse databases or clusters with ease.

Installation

To quickly get started with clickhouse-s3-etl-tools, you can install the package using the following:

pip install clickhouse-s3-etl-tools

Utilities

The ClickHouse S3 ETL Tools consist of two main utilities:

• s3_exporter: Processes data from a ClickHouse instance and exports it to an S3 bucket.

s3_exporter --ch-url-source='clickhouse+native://user:password@localhost:9000/database' \
            --s3-access-key='your_s3_access_key' \
            --s3-secret-key='your_s3_secret_key' \
            --s3-path='s3://your_bucket/path/to/data' \
            --table-name='your_table' \
            --database='your_database'

• s3_to_clickhouse_transfer: Retrieves data from an S3 bucket and transfers it to a ClickHouse instance.

s3_to_clickhouse_transfer --ch-url-destination='clickhouse+native://user:password@localhost:9000/destination_database' \
                          --s3-access-key='your_s3_access_key' \
                          --s3-secret-key='your_s3_secret_key' \
                          --s3-path='s3://your_bucket/path/to/data' \
                          --table-name='your_table' \
                          --database='your_database' \
                          --database-destination='your_destination_database' \
                          --drop-destination-table-if-exists \
                          --use-s3-cluster \
                          --cluster_name='your_directive'

Common Parameters (Applicable to both s3_exporter and s3_to_clickhouse_transfer):

• --ch-url-source (s3_exporter) / --ch-url-destination (s3_to_clickhouse_transfer):

  • Description: ClickHouse URL for either source or destination, depending on the utility.
  • Example: --ch-url-source='clickhouse+native://user:password@localhost:9000/database'

• --s3-access-key:

  • Description: Access key for the S3 bucket.
  • Example: --s3-access-key='your_s3_access_key'

• --s3-secret-key:

  • Description: Secret key for the S3 bucket.
  • Example: --s3-secret-key='your_s3_secret_key'

• --s3-path:

  • Description: Path to the data in the S3 bucket.
  • Example: --s3-path='s3://your_bucket/path/to/data'

• --table-name:

  • Description: Name of the table in ClickHouse.
  • Example: --table-name='your_table'

• --database (s3_exporter) / --database-destination (s3_to_clickhouse_transfer):

  • Description: Database in ClickHouse.
  • Example: --database='your_database'

• --batch-size:

  • Description: Batch size for data transfer (optional, default: DEFAULT_VALUE_BATCH_SIZE).
  • Example: --batch-size=100

• --log-level:

  • Description: Log level for the utility (optional, default: DEFAULT_VALUE_LOG_LEVEL).
  • Example: --log-level='DEBUG'

• --drop-destination-table-if-exists (s3_to_clickhouse_transfer only):

  • Description: Drop destination table if it exists.
  • Example: --drop-destination-table-if-exists

• --use-s3-cluster (s3_to_clickhouse_transfer only):

  • Description: Use S3 cluster for data transfer.
  • Example: --use-s3-cluster

• --cluster_name (s3_to_clickhouse_transfer only):

  • Description: Directive for cluster configuration (default: "").
  • Example: --cluster_name='your_directive'

Environment Variables:

• NUM_PARTITIONS_DROP_IN_QUERY:

  • Description: Number of partitions to drop in each query.

• MAX_TABLE_SIZE_TO_DROP_TABLE_MB:

  • Description: Maximum table size (in MB) to trigger table dropping.

• NUMB_RECONNECT_ATTEMPTS_CH:

  • Description: Number of attempts to reconnect to ClickHouse.

• MAX_PERCENTAGE_DIFF_EXTRACT:

  • Description: Maximum percentage difference for extraction.

• MAX_PERCENTAGE_DIFF_TRANSFORM:

  • Description: Maximum percentage difference for transformation.

• MAX_PARTITIONS_PER_INSERT_BLOCK:

  • Description: Maximum partitions per insert block.

• DELAY_BETWEEN_DROP_PARTITIONS_SEC:

  • Description: Delay between drop partitions (in seconds).

Example Usage with Environment Variables:

NUM_PARTITIONS_DROP_IN_QUERY=100 \
MAX_TABLE_SIZE_TO_DROP_TABLE_MB=1000 \
NUMB_RECONNECT_ATTEMPTS_CH=3 \
MAX_PERCENTAGE_DIFF_EXTRACT=1 \
MAX_PERCENTAGE_DIFF_TRANSFORM=1 \
MAX_PARTITIONS_PER_INSERT_BLOCK=500 \
DELAY_BETWEEN_DROP_PARTITIONS_SEC=10 \
s3_exporter --ch-url-source='clickhouse+native://user:password@localhost:9000/database' \
            --s3-access-key='your_s3_access_key' \
            --s3-secret-key='your_s3_secret_key' \
            --s3-path='s3://your_bucket/path/to/data' \
            --table-name='your_table' \
            --database='your_database'

ClickHouse S3 ETL Tools Documentation

File Naming Conventions

Metadata Files

Metadata files store information about the tables and their properties. These files follow the format:

• File Name Format: __metadata__[TABLE_NAME].parquet
• Example: __metadata__my_table.parquet

Full Table Export

Full table exports result in Parquet files containing all the data from the table. These files follow the format:

• File Name Format: [TABLE_NAME]_all.parquet
• Example: my_table_all.parquet

Partitioned Exports

For partitioned exports, where data is saved based on specific partition keys, files are named using the following format:

• File Name Format: [TABLE_NAME]{_partition_id}.parquet
• Example: my_table{123}.parquet

Here, {_partition_id} is replaced with the actual partition ID.

Exported File Contents

Metadata File Contents

Metadata files (__metadata__[TABLE_NAME].parquet) typically contain the following information:

• Table schema details
• Engine type
• Partitioning key
• Total number of rows

Full Table Export Contents

Files generated for full table exports ([TABLE_NAME]_all.parquet) contain the entire dataset for the specified table.

Partitioned Export Contents

Files generated for partitioned exports ([TABLE_NAME]{_partition_id}.parquet) contain data specific to the identified partition.

Dependency Tree ClickHouse

This script generates the dependency tree of ClickHouse tables and can output the result to a file or stdout. It also provides a visualization of the tree in the console.

dependency_tree_clickhouse --ch-url <ClickHouse_URL> \
                           --databases <Databases_List> \ 
                           --file-output <Output_File> \
                          [--excluded-databases <Excluded_Databases_List>] \
                          [--log-level <Log_Level>] \ 
                          [--tables <Tables_List>]  \
                          [--ignore-validation] 

Options

• --ch-url:

  • Description: ClickHouse URL.
  • Example: --ch-url='clickhouse+native://user:password@localhost:9000/database'

• --databases:

  • Description: Databases list.
  • Example: --databases='database1,database2'

• --file-output:

  • Description: File output for dict result.
  • Example: --file-output='/path/to/output_file'

• --excluded-databases:

  • Description: A list of excluded databases for the parent dependency.
  • Example: --excluded-databases='excluded_database1,excluded_database2'

• --log-level:

  • Description: Log Level (default: INFO).
  • Example: --log-level='DEBUG'

• --tables:

  • Description: Tables list.
  • Example: --tables='table1,table2'

• --ignore-validation:

  • Description: Ignore validation after fetching data.
  • Example: --ignore-validation

Example Dependency Tree

Global.root
└── test.table_cascade_null
    └── test.table_cascade_amt
        └── test._trigger_table_cascade_amt
           └── test.table_cascade_view

In this example, the dependency tree illustrates a hierarchical relationship between ClickHouse tables. The root of the tree is Global.root, and it has child tables such as test.table_cascade_null, test.table_cascade_amt, _trigger_table_cascade_amt, and test.table_cascade_view. This structure represents the dependencies between these tables, showing how they are related to each other.