Skip to main content
Version: Next

DataHubGc

DataHub Garbage Collection Source Documentation

Overview​

The DataHub Garbage Collection (GC) source is a maintenance component responsible for cleaning up various types of metadata to maintain system performance and data quality. It performs multiple cleanup tasks, each focusing on different aspects of DataHub's metadata.

Cleanup Tasks​

1. Index Cleanup​

Manages Elasticsearch indices in DataHub, particularly focusing on time-series data.

Configuration​

source:
  type: datahub-gc
  config:
    truncate_indices: true
    truncate_index_older_than_days: 30
    truncation_watch_until: 10000
    truncation_sleep_between_seconds: 30

Features​

  • Truncates old Elasticsearch indices for the following timeseries aspects:
    • DatasetOperations
    • DatasetUsageStatistics
    • ChartUsageStatistics
    • DashboardUsageStatistics
    • QueryUsageStatistics
    • Timeseries Aspects
  • Monitors truncation progress
  • Implements safe deletion with monitoring thresholds
  • Supports gradual truncation with sleep intervals

2. Expired Token Cleanup​

Manages access tokens in DataHub to maintain security and prevent token accumulation.

Configuration​

source:
  type: datahub-gc
  config:
    cleanup_expired_tokens: true

Features​

  • Automatically identifies and revokes expired access tokens
  • Processes tokens in batches for efficiency
  • Maintains system security by removing outdated credentials
  • Reports number of tokens revoked
  • Uses GraphQL API for token management

3. Data Process Cleanup​

Manages the lifecycle of data processes, jobs, and their instances (DPIs) within DataHub.

Features​

  • Cleans up Data Process Instances (DPIs) based on age and count
  • Can remove empty DataJobs and DataFlows
  • Supports both soft and hard deletion
  • Uses parallel processing for efficient cleanup
  • Maintains configurable retention policies

Configuration​

source:
  type: datahub-gc
  config:
    dataprocess_cleanup:
      enabled: true
      retention_days: 10
      keep_last_n: 5
      delete_empty_data_jobs: false
      delete_empty_data_flows: false
      hard_delete_entities: false
      batch_size: 500
      max_workers: 10
      delay: 0.25

Limitations​

  • Maximum 9000 DPIs per job for performance

4. Execution Request Cleanup​

Manages DataHub execution request records to prevent accumulation of historical execution data.

Features​

  • Maintains execution history per ingestion source
  • Preserves minimum number of recent requests
  • Removes old requests beyond retention period
  • Special handling for running/pending requests
  • Automatic cleanup of corrupted records

Configuration​

source:
  type: datahub-gc
  config:
    execution_request_cleanup:
      enabled: true
      keep_history_min_count: 10
      keep_history_max_count: 1000
      keep_history_max_days: 30
      batch_read_size: 100
      runtime_limit_seconds: 3600
      max_read_errors: 10

5. Soft-Deleted Entities Cleanup​

Manages the permanent removal of soft-deleted entities after a retention period.

Features​

  • Permanently removes soft-deleted entities after retention period
  • Handles entity references cleanup
  • Special handling for query entities
  • Supports filtering by entity type, platform, or environment
  • Concurrent processing with safety limits

Configuration​

source:
  type: datahub-gc
  config:
    soft_deleted_entities_cleanup:
      enabled: true
      retention_days: 10
      batch_size: 500
      max_workers: 10
      delay: 0.25
      entity_types: null # Optional list of entity types to clean
      platform: null # Optional platform filter
      env: null # Optional environment filter
      query: null # Optional custom query filter
      limit_entities_delete: 25000
      futures_max_at_time: 1000
      runtime_limit_seconds: 7200

Performance Considerations​

  • Concurrent processing using thread pools
  • Configurable batch sizes for optimal performance
  • Rate limiting through configurable delays
  • Maximum limits on concurrent operations

Reporting​

Each cleanup task maintains detailed reports including:

  • Number of entities processed
  • Number of entities removed
  • Errors encountered
  • Sample of affected entities
  • Runtime statistics
  • Task-specific metrics Testing

DataHubGcSource is responsible for performing garbage collection tasks on DataHub.

This source performs the following tasks:

  1. Cleans up expired tokens.
  2. Truncates Elasticsearch indices based on configuration.
  3. Cleans up data processes and soft-deleted entities if configured.

CLI based Ingestion​

Config Details​

Note that a . is used to denote nested fields in the YAML recipe.

FieldDescription
cleanup_expired_tokens
boolean
Whether to clean up expired tokens or not
Default: True
dry_run
boolean
Whether to perform a dry run or not. This is only supported for dataprocess cleanup and soft deleted entities cleanup.
Default: False
truncate_index_older_than_days
integer
Indices older than this number of days will be truncated
Default: 30
truncate_indices
boolean
Whether to truncate elasticsearch indices or not which can be safely truncated
Default: True
truncation_sleep_between_seconds
integer
Sleep between truncation monitoring.
Default: 30
truncation_watch_until
integer
Wait for truncation of indices until this number of documents are left
Default: 10000
dataprocess_cleanup
DataProcessCleanupConfig
dataprocess_cleanup.batch_size
integer
The number of entities to get in a batch from API
Default: 500
dataprocess_cleanup.delay
One of number, null
Delay between each batch
Default: 0.25
dataprocess_cleanup.delete_empty_data_flows
boolean
Whether to delete Data Flows without runs
Default: False
dataprocess_cleanup.delete_empty_data_jobs
boolean
Whether to delete Data Jobs without runs
Default: False
dataprocess_cleanup.enabled
boolean
Whether to do data process cleanup.
Default: True
dataprocess_cleanup.hard_delete_entities
boolean
Whether to hard delete entities
Default: False
dataprocess_cleanup.keep_last_n
One of integer, null
Number of latest aspects to keep
Default: 5
dataprocess_cleanup.max_workers
integer
The number of workers to use for deletion
Default: 10
dataprocess_cleanup.retention_days
One of integer, null
Number of days to retain metadata in DataHub
Default: 10
dataprocess_cleanup.aspects_to_clean
array
List of aspect names to clean up
Default: ['DataprocessInstance']
dataprocess_cleanup.aspects_to_clean.string
string
execution_request_cleanup
DatahubExecutionRequestCleanupConfig
execution_request_cleanup.batch_read_size
integer
Number of records per read operation
Default: 100
execution_request_cleanup.enabled
boolean
Global switch for this cleanup task
Default: True
execution_request_cleanup.keep_history_max_count
integer
Maximum number of execution requests to keep, per ingestion source
Default: 1000
execution_request_cleanup.keep_history_max_days
integer
Maximum number of days to keep execution requests for, per ingestion source
Default: 90
execution_request_cleanup.keep_history_min_count
integer
Minimum number of execution requests to keep, per ingestion source
Default: 10
execution_request_cleanup.limit_entities_delete
One of integer, null
Max number of execution requests to hard delete.
Default: 10000
execution_request_cleanup.max_read_errors
integer
Maximum number of read errors before aborting
Default: 10
execution_request_cleanup.runtime_limit_seconds
integer
Maximum runtime in seconds for the cleanup task
Default: 3600
soft_deleted_entities_cleanup
SoftDeletedEntitiesCleanupConfig
soft_deleted_entities_cleanup.batch_size
integer
The number of entities to get in a batch from GraphQL
Default: 500
soft_deleted_entities_cleanup.delay
One of number, null
Delay between each batch
Default: 0.25
soft_deleted_entities_cleanup.enabled
boolean
Whether to do soft deletion cleanup.
Default: True
soft_deleted_entities_cleanup.futures_max_at_time
integer
Max number of futures to have at a time.
Default: 1000
soft_deleted_entities_cleanup.limit_entities_delete
One of integer, null
Max number of entities to delete.
Default: 25000
soft_deleted_entities_cleanup.max_workers
integer
The number of workers to use for deletion
Default: 10
soft_deleted_entities_cleanup.platform
One of string, null
Platform to cleanup
Default: None
soft_deleted_entities_cleanup.query
One of string, null
Query to filter entities
Default: None
soft_deleted_entities_cleanup.retention_days
integer
Number of days to retain metadata in DataHub
Default: 10
soft_deleted_entities_cleanup.runtime_limit_seconds
integer
Runtime limit in seconds
Default: 7200
soft_deleted_entities_cleanup.env
One of string, null
Environment to cleanup
Default: None
soft_deleted_entities_cleanup.entity_types
One of array, null
List of entity types to cleanup
Default: ['dataset', 'dashboard', 'chart', 'mlmodel', 'mlmo...
soft_deleted_entities_cleanup.entity_types.string
string

Code Coordinates​

  • Class Name: datahub.ingestion.source.gc.datahub_gc.DataHubGcSource
  • Browse on GitHub

Questions

If you've got any questions on configuring ingestion for DataHubGc, feel free to ping us on our Slack.