tpakeman / concurrent-content-validation

concurrent-content-validation

Basics

• Initial implementation of an idea to concurrently run multiple Looker content validator runs, each one scanning 1/nth of the available content
• The full idea is below (with an indication of if it is implemented):
  • Parse all of the Looker folders ✅
    • Preserve parent-child relationships ✅
    • Identify enclosed dashboards and Looks ✅
    • Count the total number of enclosed queries ✅
  • Split the folders into n slices of ~equal query volume ✅ needs checking
    • Capture their content_metadata_ids ✅
      • Ensure each exists in a path containing all parent folder IDs for inheritance ✅ needs checking
  • Create n users with the develop permission and no folder access ❌
    • Give each user access to one of the n slices of content ❌
    • Run the validator for each user ❌

Bugs + missing features

• There are still a number of SDK errors being raised by the content validator
• This does not yet handle user creation or deletion
• This only works with a closed system enabled
• This does not scan LookML dashboards (but neither does the content validator)
• This only imperfectly divides queries into 1/n fractions (and needs to be tested)

Usage

• Save a looker.ini file in the root directory
• Run main.py and name a section in the ini file to auth into
  • e.g. /path/to/main.py Section Name {-arg foo --other-arg bar}
• Use the following cli arguments:
• options:
  • --user USER, -u USER The ID of the user to impersonate for validator runs
  • --timeout TIMEOUT, -t TIMEOUT Set a max timeout for content validator runs
  • --fractions FRACTIONS, -f FRACTIONS How many equal sized fractions of the content should be validated in each run
  • --iterations ITERATIONS, -i ITERATIONS How many times should each validator run execute
  • --silent, -s Flag to suppress progress printing as the folder tree is scanned
  • --create_users, -c Flag to create users rather than use a named user ❌ not implemented