request help: Will limit-req support nodelay and rate in minute

[nanamikon]

Issue description

Will limit-req support nodelay and rate in minute?
If not, I find nodelay can be implemented by adding condition, but resty.limit.req only support rate in second, how can I implemnt rate in minute?

Environment

Request help without environment information will be ignored or closed.

• apisix version (cmd: apisix version): 2.5
• OS (cmd: uname -a): ubuntu
• OpenResty / Nginx version (cmd: nginx -V or openresty -V):
• etcd version, if have (cmd: run curl http://127.0.0.1:9090/v1/server_info to get the info from server-info API):
• apisix-dashboard version, if have:
• luarocks version, if the issue is about installation (cmd: luarocks --version):

[spacewander]

I find nodelay can be implemented by adding condition

PR is welcome!

how can I implemnt rate in minute

Divide the rate by 60 is already enough? IMHO, there is no need to add a field. Since Nginx also does the same way: https://github.com/nginx/nginx/blob/5eadaf69e394c030056e4190d86dae0262f8617c/src/http/modules/ngx_http_limit_req_module.c#L913

[nanamikon]

In my company, many users set 200 r/m on remote_address or even lower, If it divided by 60, 1r/s or 2r/s will be reached easily, because modern browser support http2, so one user visit one nignx over 1r/s is very common.

Now, we transfer this conf to apisix, some of them reach limit and delay request online。

I think we use nodelay in nginx conf while apisix only support delay cause this happen， because both of them reach the rate limit, but not over burst. I will test it

The rate is specified in requests per second (r/s). If a rate of less than one request per second is desired, it is specified in request per minute (r/m). For example, half-request per second is 30r/m.

ctx->rate = rate * 1000 / scale;

Maybe multiply by 1000 is the key to solve the decimal point ？

[spacewander]

In my company, many users set 200 r/m on remote_address or even lower, If it divided by 60, 1r/s or 2r/s will be reached easily,

It's how Nginx works too. Note that the r/m is the speed rate, it doesn't mean you can have 200 requests in a minute, it means you can't be faster than 200r/m. Drive a car in 36km/h is equal to 10m/s.

because modern browser support http2, so one user visit one nignx over 1r/s is very common.

In this case we need a good burst and/or use delay.

ctx->rate = rate * 1000 / scale;

resty.limit.req does the same thing, see https://github.com/openresty/lua-resty-limit-traffic/blob/56b9a35a69117ed0e8a2e3b9a0866c426403e1b9/lib/resty/limit/req.lua#L144

[nanamikon]

Thanks for reply， after adding nodelay， we solve the online problem

resty.limit.req does the same thing, see https://github.com/openresty/lua-resty-limit-traffic/blob/56b9a35a69117ed0e8a2e3b9a0866c426403e1b9/lib/resty/limit/req.lua#L144

So mayby rate should be float, but I think it is not the key point

In this case we need a good burst and/or use delay.

I want to submit a pr to add nodelay function, a new param as follow

Name	Type	Requirement	Default
nodelay	boolean	optional	false

Is it ok?

[spacewander]

LGTM