Prisma

<@UU7HJTD1S> are you doing anything differently?

Ive honesty found kubernetes too complex for my use causes atm. Running Prisma behind a load balancer in Elastic Beanstalk on AWS and it's working pretty well and does not take too long to spin up new instances. 

<@UU7FS3EM9> how are you scaling your DB along with prisma?

<@UU7FS3EM9> <@UU7HJTD1S> I’m running 20 prisma-prod containers on ecs with database connections upped to 100 each, 2 cpus, 4gb ram, with an external rabbitmq server and every few minutes a few containers crash from subscriptions and respawn.

We use subscriptions heavily, more than anything else.

how are you able to set the db connections?

```databases:
                    default:
                      connector: ${DbConnector}
                      host: ${DbHost}
                      port: ${DbPort}
                      user: ${DbUser}
                      password: ${DbPassword}
                      migrations: true
                      rawAccess: true
                      connectionLimit: 100```

How is that any different that queuesize?

I’m really unsure what queuesize does. but we increased it to 20k in hopes it would help and it did nothing but make prisma fall over faster

so the thing about prisma is that it doesn't utilize any read endpoints...

so using read replicas is basically useless, right?

not only that but prisma 1 is maintainance only, so don’t expect much from it.

how many users do you currently have Ethan?

I'm load testing for something related to auth tokens, and it crashes at 280

It’s hard to say, because we don’t have good app insights, but we are trying to support 2k concurrent but are likely falling over around 200~

and that’s with all of the optimization / scaling suggestions in place

was wondering if you are using alb to auto scale your ecs containers

do you know what the connection Limit default is?

I am using fargate and they are behind a load balancer