Multi Container options and High Availability Possibilities

[Aevrin]

I was wondering what the options were having freshrss in any sort of high availability without using swarm and/or kubernetes. I was thinking either if I use sqlite stored on a shared filesystem (either local, NFS, or something fancy like ceph), or if using mysql and copying the volume data, then they should all reference the same DB.
When testing using sqlite on a local bind mount, it seems to work really well. I configured it in one container, then they all work, but I suspect this isn't supported.

version: "3.3"
  services:
    freshrss-1:
      container_name: Freshrss-Compose-1
      hostname: freshrss-compose-1
      image: freshrss/freshrss:latest
    restart: always
    volumes:
      - /Data/RSSS/data:/var/www/FreshRSS/data/
      - /Data/RSSS/extensions:/var/www/FreshRSS/extensions
    ports:
      - 8081:80
  freshrss-2:
    container_name: Freshrss-Compose-2
    hostname: freshrss-compose-2
    image: freshrss/freshrss:latest
    restart: always
    volumes:
      - /Data/RSSS/data:/var/www/FreshRSS/data/
      - /Data/RSSS/extensions:/var/www/FreshRSS/extensions
    ports:
      - 8082:80
  freshrss-3:
    container_name: Freshrss-Compose-3
    hostname: freshrss-compose-3
    image: freshrss/freshrss:latest
    restart: always
    volumes:
      - /Data/RSSS/data:/var/www/FreshRSS/data/
      - /Data/RSSS/extensions:/var/www/FreshRSS/extensions
    ports:
      - 8083:80

[Frenzie]

I believe SQLite should work fine with GlusterFS but I haven't tried it.

[Aevrin]

Gluster falls into my something fancy like ceph bucket. So I may have to figure it out, but right now I'm trying to avoid it. But it's refreshing to hear that it should work as needed.

[Alkarex]

See also #5748 and help welcome there to mature and test

[Aevrin]

Testing in the long term I get random errors that a refresh fixes, but it's more of a pain. I'll probably switch back to a single instance unless someone wants logs from my non-supported setup.

[Alkarex]

It is not something I will personally be looking into in the very near future, but other people may be interested, so if you have some relevant logs to share, please do

[Aevrin]

This is what I get:
SQL error FreshRSS_EntryDAOSQLite::commitNewEntries["HY000",5,"database is locked"]

I have the cron jobs for each container so they shouldn't interfere. I also notice periodic issues where it has the wrong rid string for the container I connected to as I'm simply letting dns load balance and I've been shifted between them with no control. It's possible if that are removed it wouldn't provide some of the errors.

[Alkarex]

I believe with SQLite, you need to make sure that there is only one instance at a time for a given user.
It could be interesting to try with e.g. PostgreSQL as database.

[Aevrin]

A single postgres instance, multiple replicated, or multiple using the same storage on NFS?

[Alkarex]

I believe even a single PostgreSQL instance is less likely to be susceptible of frequent concurrency locks than SQLite for this specific setup