I hope this post is not too off topic. I thought that it would be nice to see the address of all the small self-hosted instances of Lemmy (1~5 users).

    • tyfi@wirebase.orgEnglish
      3·
      1 year ago

      Is there a way to host with high availability? Or is that a kubernetes feature?

      • kelvie@lemmy.caEnglish
        3·
        1 year ago

        K8s is just a huge abstraction over your clusters, the real question is if the software/containers support HA.

        • tyfi@wirebase.orgEnglish
          1·
          1 year ago

          I’ve been meaning to test it for a while now, but have just been running VMs/Docker. Will check it out.

      • Philip@endlesstalk.orgEnglish
        1·
        1 year ago

        You can definitely have high availibillity without kubernetes, but its easier(For me atleast) with kubernetes.

          • Philip@endlesstalk.orgEnglish
            1·
            1 year ago

            For container orchestration, which is mostly what k8s provides, then you could use docker swarm or nomad. You could use docker-compose with multiple replicas of the wanted container + a load balancer to divide the load.

            In general I don’t think k8s/k3s is needed for hosting lemmy yet, but since I have a setup for k3s, it is easier for me to use it.

    • blazarious@mylem.meEnglish
      1·
      1 year ago

      how do you handle the sled state for pictrs with 2 nodes? I’ve been having some trouble with it.

      • Philip@endlesstalk.orgEnglish
        2·
        1 year ago

        I have only 1 container of pictrs running(with no scaling) and are using longhorn for storage, so if the pictrs container switches node, then longhorn handles it for me.

        • blazarious@mylem.meEnglish
          1·
          1 year ago

          I see, thanks. What volume(s) are you persisting that way exactly? I mean the internal path that pictrs is using.

          • Philip@endlesstalk.orgEnglish
            2·
            1 year ago

            The internal path, I’m persisting is /mnt, but I also use an older version of pictrs(0.3.1). Think the newer version uses a different path.

            I also needed to add the following for the pictrs container to work correctly.

              securityContext:
    runAsUser: 991
    runAsGroup: 991
    fsGroup: 991