https://www.prisma.io/ logo
Join Slack
Powered by
# orm-help
  • e

    Ethan Pursley

    03/20/2020, 12:29 AM
    the default is 10
  • a

    Alan

    03/20/2020, 12:29 AM
    oh jeez
  • a

    Alan

    03/20/2020, 12:30 AM
    ok that might help me
  • a

    Alan

    03/20/2020, 12:30 AM
    thanks
  • a

    Alan

    03/20/2020, 12:32 AM
    but yeah did you set an alb to autoscale ethan?
  • e

    Ethan Pursley

    03/20/2020, 12:33 AM
    Not yet, i manually ramp up right now. I have no real metrics to scale on at the moment. The prisma cpu/memory usage is around 10% of the total but prisma keeps crashing and spinning back up
  • a

    Alan

    03/20/2020, 12:34 AM
    hm
  • a

    Alan

    03/20/2020, 12:34 AM
    could it be your limits?
  • e

    Ethan Pursley

    03/20/2020, 12:34 AM
    I’m not sure, all the limits are scaled up and likely not even being hit
  • e

    Ethan Pursley

    03/20/2020, 12:35 AM
    the errors that i see all come from the scala pubsub stuff that is used inside prisma
  • a

    Alan

    03/20/2020, 12:35 AM
    what does it say?
  • e

    Ethan Pursley

    03/20/2020, 12:35 AM
    we are a heavy user of subscriptions that are not well supported in primsa 1 and not at all in prisma 2
  • a

    Alan

    03/20/2020, 12:35 AM
    how many rabbitmq nodes are you running?
  • e

    Ethan Pursley

    03/20/2020, 12:36 AM
    just one large one
  • a

    Alan

    03/20/2020, 12:36 AM
    i know you have it externally hosted but do you have any idea on the amount of nodes for the queue?
  • a

    Alan

    03/20/2020, 12:36 AM
    oh
  • a

    Alan

    03/20/2020, 12:36 AM
    i had to bump the rabbitmq workers to 40
  • e

    Ethan Pursley

    03/20/2020, 12:37 AM
    Copy code
    [INFO] [03/18/2020 21:45:34.837] [single-server-akka.actor.default-dispatcher-7] [<akka://single-server/user/$3>] Message [akka.actor.Status$Success] without sender to Actor[<akka://single-server/user/$3#-2128321766>] was not delivered. [10] dead letters encountered, no more dead letters will be logged. This logging can be turned off or adjusted with configuration settings 'akka.log-dead-letters' and 'akka.log-dead-letters-during-shutdown'.
  • e

    Ethan Pursley

    03/20/2020, 12:37 AM
    Well i have one rabbitmq server, right now i’m just running a bitnami instance with whatever defaults it sets for rabbitmq
  • e

    Ethan Pursley

    03/20/2020, 12:37 AM
    it was sort of a ‘hail mary’ rush to try and get it to scale
  • e

    Ethan Pursley

    03/20/2020, 12:38 AM
    but it’s still not working, hints why i’m here trying to ask for advice haha
  • e

    Ethan Pursley

    03/20/2020, 12:38 AM
    My team and I were on a call with one of the prisma engineers yesterday and he basically told us if we need subscriptions, we should switch to something like hasura.
  • e

    Ethan Pursley

    03/20/2020, 12:38 AM
    you may want to do the same if you need them haha.
  • a

    Alan

    03/20/2020, 12:46 AM
    we have considered it but we have built around prisma a bit for it to be used as our ORM for our other services too
  • e

    Ethan Pursley

    03/20/2020, 12:50 AM
    Yeah, we have too, so it’s hard to move away from it at this time. Which sucks because we can’t get the damn thing to scale, even with throwing tons of money at it
  • a

    Alan

    03/20/2020, 12:57 AM
    well, setting the connection limit didn't seem to help for me either
  • a

    Alan

    03/20/2020, 1:05 AM
    at this point i'm not sure if it's worth it to migrate to 2. i'm sure moving away from the per pod client is a better model but
  • a

    Alan

    03/20/2020, 1:06 AM
    there's a lot of things like read replicas and etc that are yet to be included
  • a

    Alan

    03/20/2020, 1:14 AM
    @Ethan Pursley are you running Prisma Pro or Premium?
  • a

    Alan

    03/20/2020, 1:14 AM
    or just the open source implementation?
1...355356357...637Latest