Running v9 in parallel with v8, sharing same database

Issue Summary

Running v9 in parallel with v8

Technical details:

  • Redash Version: v8 latest, v9 preview
  • Browser/OS: n/a
  • How did you install Redash: AWS ECS

I have finally gotten around to trying out the v9 release, and I’d like to get it in the hands of my users now to test. What I’d like to do is configure my v9 install to share the existing v8 database, but before I do so, I wanted to see if v9 introduced any database changes that would prevent this, or if generally this is a Bad Idea ™. The database would be the only thing v8 and v9 installs would share, and I’d really like to avoid reconfiguring all 50+ datasources and redoing all ACLs for a v9 standalone database.


Those are the migrations. There are a few changes between v8 and v9.

I don’t think anyone would recommend what you describe. Lest you fubar your entire setup (which is more hassle than reconfiguring data sources and ACLs by hand :wink:). That said, if you’re running Docker it should be trivial to fork your V8 instance and upgrade it to V9 for testing. You shouldn’t need to manually reconfigure anything.

@k4s1m thanks for the information! I’ll mirror my existing v8 DB for v9 testing in that case.

Do you (or anyone) happen to know when the repo will be updated for v9? While I read through the changelog for v9 and updated as suggested, it didn’t make much sense, and I ended up with the following configurations for my Fargate ECS tasks:

  • Front end service running tasks for server
  • Scheduler service running ONE task for scheduler
  • Scheduled worker service tasks for worker set for queues periodic,emails,default,scheduled_queries,schemas
  • Adhoc query service running tasks for worker set for queues queries

The one gotcha I had when I updated to v9 is that my services are typically configured to run two copies of the task for multi-AZ deployment, and the scheduler command on the second task would fail with an error that the RQ scheduler was already running, so I had to reduce the task count for that service to 1. I don’t know if this is something the updates to the scheduler service will address, but it’d be nice to be able to run multiple copies of that task.