# R2 Rate Limiting

## Context

- 2 workers
- api.nft.storage
- api.web3.storage
- 3 buckets used by both workers with given key formats
- carpark-prod-0 => `${carCid}/${carCid}.car`
- satnav-prod-0 => `${carCid}/${carCid}.car.idx`
- dudewhere-prod-0 => `${rootCid}/${carCid}`
- `/upload` and `/car` routes trigger writes to all these 3 buckets

## Monitoring



### Observations

- Pink user doing a quite significant number of requests and no significant errors happening
- Green user starts performing a large number of requests, starting triggering a lot of errors in our API for R2 rate limiting
- not only for this user, but other users are also affected
- also, other project in same account (nft.storage) also started having rate limit errors
- Error received in worker seems to indicate rate limit is based on the same object

## Questions

1. What is the actual rate limit in terms of requests/time?
2. How is rate limit applied?
- Is it actually per object key, or any type of prefix?
- Is this accounter per worker/bucket, or can it be at account level?
- Looks like same prefix key in multiple buckets can have an effect...
3. What happens when we get over the limit and get rate limited?
- Does this affect other buckets?
- Does it even affect the entire account?

## What we need

- We can work around a rate limit per key in a single bucket, but not if this affects other users/buckets. It is super unlikely that different users will generate same keys