Conversation

Notices

  1. @moonman !loadaverage is running on an HDD and 4G of RAM %).
    We don't have Qvitter though, this may be the key resource-eater for the GNU social installations out there.

    Sunday, 06-Jan-19 20:12:03 UTC from loadaverage.org
    1. @xrevan86 @moonman HLA has never really had performance problems, even with Qvitter, and LA's VM was copied from this instances (same specs).

      It really comes down to optimising your database.  Database use is the primary bottleneck.  This is why optimising the joins like you did Alexei gave you the gains you got.  You're hitting the database 100,000s of time in a minute if you federate with much people at all.

      Sunday, 06-Jan-19 20:40:37 UTC from community.highlandarrow.com
      1. @xrevan86 @moonman When I have had problems has always been when I'm livehacking on things, and the responsible party in that case is always the same thing.  Me.

        Sunday, 06-Jan-19 20:43:07 UTC from community.highlandarrow.com
      2. Using the redis queue is the huge improvement it is, specifically because it greatly reduces the time for common queries to static information.

        Sunday, 06-Jan-19 20:46:42 UTC from community.highlandarrow.com
        1. That said Redis has always been unreliable in the sense that every update breaks banana, so I might consider alternatives, or homebrew a caching thing, when I get that far.

          Sunday, 06-Jan-19 20:49:15 UTC from community.highlandarrow.com
          1. @maiyannah This raises the question: does the redis queue use drive storage?

            Sunday, 06-Jan-19 23:57:37 UTC from loadaverage.org
            1. @xrevan86 No, it is a memory-based cache unless you tell it specifically to page to the drive (which you can do, but it is not so by default)

              Monday, 07-Jan-19 00:39:13 UTC from community.highlandarrow.com