Rate limiting

[DavidM-D]

Create simple configurable limits on the number of requests that a given SDK user can make.

Aim of the game is to stop them bringing down the MPC recovery service, not that we accurately bill them.

Tasks

Preview Give feedback

1. Decide what CORs we want to allow mpc-recovery#186
   Decision Emerging Tech Near BOS +3
   

[volovyks]

Looks like GCP has what we want out of the box.
https://cloud.google.com/armor/docs/rate-limiting-overview

I think throttling based on ip/domain is what we need. Each partner should have their own domain since it's a separate Firebase app.

We can add protection from individual users too.

cc @itegulov do you think it's the best way to do it?

[volovyks]

My current plan is to set up the load balancer using Terraform and Google Armor.

[volovyks]

Details for ratelimiting are here: #289 (comment)
and other comments.

[kmaus-near]

Hey guys, I've got some URLs for you to use that put MPC behind Kong for rate limiting policies to take effect:

Mainnet/Prod: https://near-mpc-recovery-mainnet.api.pagoda.co

Testnet/Dev: https://mpc-recovery-leader-testnet.dev.api.pagoda.co

Nothing special required to use these URLs, but I think once we fully transition to use these, we should switch the default CloudRun URLs to internal only so we don't circumvent the Kong rate limit/loadbalancer.

[volovyks]

@kmaus-near seems like you are the best person to make this change. Can you help us with that? Let's finish with it for 100%.

[volovyks]

@kmaus-near You will probably need to sync with @esaminu
cc @itegulov

[kmaus-near]

Update from my side of this, in order to make those cloudrun auto generated URLs private, I'll be using an internal loadbalancer so the Kong proxy can still reach the services. Once I finish that LB and test I'll reach out to @esaminu and make sure things are good.

[kmaus-near]

We'll have to somewhat work together on this, I created a branch @ https://github.com/near/mpc-recovery/tree/kmaus-near/add-internal-lb/infra

Take a look at this and lmk if something like this would work for your env. If not we might want to put prod and dev TF code into their own directories so we can have a bit of separation going on as long as it doesn't break any of your workflows.

[itegulov]

@kmaus-near yeah I think this is reasonable. We can create some testnet/mainnet separation if necessary. Shouldn't be too big of a deal.

[volovyks]

Added on a load balancer level.