Cache or some temporal memory?

[jesus-hergueta]

Hi.

I'm opening this question because I'm having a problem with some dashboard I have.

Basically I have a Docker compose stack where I have a container running Datasette and another one with a Cron that runs a pipeline. The cron every night updates the .db file that Datasette is using. This file is in a bind folder configured in both containers, and the folder is in the host OS.

Everything was working fine until 2 weeks ago.

I have a dashboard built with the plugin datasette-dashboards and when the cron ran and update the DB file the queries in this Dahsboard start to return incorrect data and even zero results.

The problem is solved with a restart of the containers (I don't even have to rebuild them).

There is any memory that must to be reseted or some cache that can help me fix this problem?

I've already check the docs and find Immutable mode and Datasette inspect or Caching but I don't have any of that in my configurations (maybe is loading this configurations by default?).

[Muguett-DBY]

Does the pipeline update the existing database with SQL, or rebuild and replace the .db file?

I checked both with Datasette 0.65.5 on Linux (run). A committed SQL update became visible to the existing read pool. Replacing the file left that pool reading old rows, while a fresh SQLite connection and a fresh Datasette pool saw the replacement. Both cases used mutable mode.

If your cron replaces the file, close the writer and stop Datasette before swapping it, then start Datasette again. Alternatively, update the existing database through committed SQLite transactions. SQLite explains the problems with renaming an open database.

Also compare the dashboard's query with the same SELECT in Datasette's SQL page. If that result is current, investigate the dashboard/browser caching separately.

I haven't tested your Docker/dashboard stack. Checked with Codex.