Skip to content

Looker Slowness due to Queuing

Slowness usually means many queries hit Snowflake at once. A dashboard with many tiles sends one query per tile.

Snowflake runs at most 8 queries in parallel, and in practice 3-4 for complex Looker queries, so the rest queue. The MTU deep dive dashboard on the Finance explore sends about 27 queries at once, and other users on the same explore compete for the same resources.

Results are cached for a day, so it mostly shows on the first load of the day or with Clear cache and refresh.

Workarounds

  • Remove tiles that are not needed
  • Split the dashboard by business context
  • Schedule an early morning run to warm the cache

If none fits, discuss the business need and options such as pre-calculating metrics.