Looker Performance Optimization - Episode 2

Key Concepts:

• Looker instance performance optimization
• Henry tool (command-line tool for Looker)
• Database connections
• Unused projects and models
• Slow dashboards and explorers
• Legacy features
• LookML errors

Using Henry to Assess Looker Instance Health

Mike DeAngelo, a developer relations engineer at Google, continues the series on optimizing Looker instance performance by focusing on the Henry tool.

1. Running the Pulse Command:

• The bin/henry pulse --section=sandboxcl command is used to run a series of checks to assess the overall health of the Looker system.
• The command execution time is approximately 5-10 minutes, depending on the instance size.
• The pulse command consists of six checks.

2. Connection Checks and Query Counts:

• The first check verifies the database connections.
• It checks if Looker can successfully connect to the underlying databases.
• It retrieves the query count for recent queries against those databases.
• The output displays the status and any error messages.
• Error messages indicate that associated models are likely not being used.
• Zero queries also suggest that the associated models are not in use.
• This is the initial step in identifying unused projects and models.

3. Dashboard Performance Checks:

• Henry checks for dashboards with queries slower than 30 seconds in the last seven days.
• This identifies dashboards that need database optimization.
• Dashboards with errors in the last seven days are also checked.
• These errors indicate problems with the LookML code that need to be addressed.

4. Explorer Performance Checks:

• The slowest explorers are identified.
• These explorers should be examined for potential database connection and query optimizations.

5. Legacy Feature Checks:

• Henry checks for enabled legacy features.

6. Reducing Unnecessary Projects and Models:

• The primary key to performance improvement is reducing the number of unnecessary projects and models.
• Unused database connections provide clues about potentially removable projects and models.
• Caution is advised, as some items might be used infrequently for auditing or other purposes.
• Additional research may be needed to confirm if an item can be safely deleted.

7. Next Steps:

• The next session will delve deeper into the Henry tool to find more details about items that can be removed and optimized.

Notable Quotes:

• "The first key to performance improvement is reducing the number of unnecessary projects and models." - Mike DeAngelo

Synthesis/Conclusion:

The Henry tool's pulse command provides a comprehensive initial assessment of a Looker instance's health. By checking database connections, identifying slow dashboards and explorers, and highlighting unused projects and models, Henry helps pinpoint areas for optimization. The process involves a combination of automated checks and manual investigation to ensure that only truly unnecessary components are removed, leading to improved Looker instance performance.