Hi all, a shout-out for assistance. I’m considering hosting a Lemmy instance (assuming I can pass the wife test on costs) and I’m looking for some guidance on specs.

Can anyone who’s currently hosting an instance (or who knows the inner workings of one) please reply with:

  • specs on the hardware / VPS that’s hosting your instance
  • how many users / posts that’s supporting
  • what the system load looks like with the above
  • if locally hosting, the type of bandwidth requirements you’re seeing

I previously posted this in the wrong community, and one of the responses asked how many users I’m expecting. To preemptively answer - I don’t know. I’m just trying to get an idea of relative sizing.

Thank you!!

  • Ducks@ducks.dev
    link
    fedilink
    English
    arrow-up
    3
    ·
    1 year ago

    It’s pretty lightweight. I’ve given each container 1/3Gi of memory and 1CPU limit with low requests. Utilizing kubernetes HPA to scale containers under load up to 4 replicas. It only scales when a user takes large actions (subscribing to hundreds of newto. The instance communities at once). But once the initial federation begins it seems to quickly scale back down. The biggest bottleneck is pictrs since it is stateful.

    So far the database and pictrs is only about 2Gi of storage but I’ve allocated 25Gi to each since I have a lot to spare at the moment.

    I have to play with the HPA more since I’m not happy yet with my settings. I have 2 users and 1 bot on my instance.

    I’d like to start contributing to Lemmy’s codebase so I wanted to host my own instance to learn the inner workings.

    My postgres is a single replica at the moment but I may scale that if it begins to bottleneck.