Metabase dashboards stay blank for ~20 minutes after hourly Redshift table reload

I’m facing an issue with Metabase + Amazon Redshift related to data freshness and dashboard availability.

I have an hourly ETL job that drops and recreates a table in Redshift, then reloads the data.
The load itself takes around 5 minutes.

During this window, the Metabase chart that depends on this table goes blank, which is expected.
However, the real problem is that even after the load finishes, Metabase keeps showing the chart as empty for ~15 additional minutes. So instead of ~5 minutes of downtime, the dashboard takes ~20 minutes to recover.

Additional details:

  • The Redshift table is dropped and recreated every hour

  • Data is fully available in Redshift immediately after the load finishes

  • Metabase seems slow to “see” the table/data again

  • This looks related to caching, metadata sync, or schema refresh behavior

  • Dashboard auto-refresh is enabled every 5 minutes, but it does not seem to work consistently for all users

  • I enabled auto-refresh, but other users still see blank charts even after the data is already available in Redshift

Questions:

  1. Does Metabase cache table metadata or query results in a way that causes delays after DROP TABLE / CREATE TABLE?

  2. Is dashboard auto-refresh expected to force re-execution for all users, or can cached results / metadata still be served?

you probably have caching enabled

So you mean when cache is enabled, the auto-refresh doesn't work?

If caching is enabled Metabase will pull the queries (cached ones) from the cache till it expires