Migrating from docker to non-docker: Is Redis database re-generatable or mergeable?

[lo48576]

I'm migrating from docker-compose deployment to non-docker, and it almost succeeded... without Redis part.

Current docker-compose.yml on production branch is using two Redis 7.2.1 servers for redis_activity and redis_broker, but the latest package on Debian 12 (bookworm) is Redis 7.0.15 and the official documentation share the localhost for two purposes.

So I googled a bit but cannot find the ways to:

• moving the Redis data from newer (7.2.1) DB to older (7.0.15) DB, and
• merging two databases (redis_activity and redis_broker in docker compose) into one (redis.service on Debian).

So my question is, how this migration can be done?
Is there any ways to successfully migrate Redis data?
Or the data on Redis can be re-generated from the data on Postgresql?

[lo48576]

Updating Your Instance

Feeds for each user are stored in Redis. To re-populate a stream, use the management command:

Oops I overlooked this, but is this enough for filling empty Redis database?
I assume so, but I'm not exactly sure what the two Redis servers do and whether it is safe to resetting them.

[mouse-reeve]

Redis activity can be completely regenerated from nothing using the script you mention. Redis broker however stores data for celery, and can’t be rebuilt. It handles executing longer running background scripts and communicating with external instances. If there are no pending celery tasks, deleting the data won’t cause any disruption to service, but that’s not quite the same thing.

[lo48576]

Thank you.

It handles executing longer running background scripts and communicating with external instances. If there are no pending celery tasks,

Would it be (relatively) safe to ignore some "longer running background scripts"?
My instance is a personal single-user almost-no-followers server, so one-time inconsistency on federation is acceptable for me.
If the bad effect of "disruption" does not persist locally, I'm OK to regard the migration as completed (without checking whether some tasks had been pending).