Ratelimits, Queues and other questions.

[silver junco]

Hi, I have a few questions about the ratelimit handing, queues and invalid request bucket code.

First of all, do interactions go into the same queue as other requests such as getting a channel. If they do how can this be circumvented as those requests do not fall under the same global ratelimit as mentioned in the docs.

Secondly, what is the invalid request bucket, what is the point of it? Will it automatically prevent making invalid requests or is it to just store endpoints? I currently have a system that checks if a resource, such as channels/roles, is invalid and block the request however I am wondering if there is a better way to do this utilising the invalid request bucket.

[vague marsh]

do interactions go into the same queue as other requests such as getting a channel.
no, i belive they get in a different queue based on the interaction id

what is the invalid request bucket, what is the point of it?
Avoid the Cloudflare ban. Discord will do ban the ip IF their api responds 10k times with 400, 429* or 403

ill it automatically prevent making invalid requests or is it to just store endpoints?
If you reach 10k it will prevent any request to discord until the headers it got tells it that are safe and unbanned

*: Shared ratelimits do not count, however most stuff is token / ip based

[silver junco]

So basically the invalid request is just counted in the bucket and sits near the front of the makeRequest function only intervening when it reaches 10k requests however it will not actually stop requests that could still be invalid.

Do 404 requests not count as invalid? I feel like it should.

If I wanted to implement something that would intervene in a more organised approach to my current method caching invalid channels, is there a way I can check this invalidRequests queue or is it internal only

[vague marsh]

they don't afaik (the code does not take them into account either)

[vague marsh]

you could override sendRequest in the rest manager and so do the fetch call yourself
however i don't know if that would be the appropriate solution in this case

[silver junco]

We already proxy the rest manager as it is so in the seperate rest process we just put a function that checks if its invalid infront of sending the request further to discordeno

But what I would like to do is match requests coming in to ones found in the invalidRequests queue

so I can mark it as invalid and not actually make the request again

[vague marsh]

invalidRequests queue
the issue is that it isn't a queue
it is a glorified counter if we call it in this way

[silver junco]

Okay that makes sense

[vague marsh]

we only store the number of times an invalid request got sent (and discord responded with one of those status code)

• other metadata for the count (max, the ammount to be safe (1) the interval and so on)

[silver junco]

Okay so really I need to have my own implementation just based off of errors

[whole reef]

Do 404 requests not count as invalid?
They don't but too many 404s isn't nice either, like there's an undocumented limit for how many webhooks you can get 404 for afaik so best to avoid 404s as well

[silver junco]

Gotcha

[whole reef]

For what?