Handling concurrent requests in lua-resty-redis module

shashwat.shalvi.y14 · 2018-06-02T18:33:37+00:00

Probably not documented anywhere.There's generally pretty active chatter on relevant Github projects. As for larger real-time chat applications, that gen...

Handling concurrent requests in lua-resty-redis module

shashwat.shalvi.y14

Hi all,

I have recently started using lua and openresty. I am trying to build an API gateway through openresty. One of the issues I am facing is handling concurrent requests in lua-resty-redis. Whenever the nginx server receives a request (request B) after a blocking request (request A) a new redis connection is created for the latest request( B ). I am using set_keepalive method of resty.redis to keep the pool of 10 connections alive for 100 seconds. What I am not able to deduce is when the connection from this pool will be used and when will be a new connection created. As of now whenever a new request comes after a blocking request a new connection is being created. This connection might be from the pool or it might be a new one.

I used get_reused_times method to check whether request B was using the previous connection. It turns out it is not.

Also if some one could explain me how it is related to Worker connections I will be thankful.

All in all I just want to know when a new connection would be created and when a connection from the pool would be used.

Thanks in advance.

P.S : This the first time I am posting a message in an opensource community, so please don't mind anything.

Shashwat.

rpaprocki

Hi,

On Sat, Jun 2, 2018 at 3:33 AM, Shashwat Shalvi <shashwat....@lnmiit.ac.in> wrote:

Hi all,

I have recently started using lua and openresty. I am trying to build an API gateway through openresty. One of the issues I am facing is handling concurrent requests in lua-resty-redis. Whenever the nginx server receives a request (request B) after a blocking request (request A) a new redis connection is created for the latest request( B ). I am using set_keepalive method of resty.redis to keep the pool of 10 connections alive for 100 seconds. What I am not able to deduce is when the connection from this pool will be used and when will be a new connection created. As of now whenever a new request comes after a blocking request a new connection is being created. This connection might be from the pool or it might be a new one.

I used get_reused_times method to check whether request B was using the previous connection. It turns out it is not.

Also if some one could explain me how it is related to Worker connections I will be thankful.

All in all I just want to know when a new connection would be created and when a connection from the pool would be used.

Overall, you may want to have a read of https://github.com/openresty/lua-nginx-module#tcpsocksetkeepalive, as this describes the behavior of connection pooling fairly well. In general, connections are put into the connection pool immediately when caling sock:setkeepalive() (assuming the call was successful), and when a new connection is instantiated via https://github.com/openresty/lua-nginx-module#tcpsockconnect or https://github.com/openresty/lua-nginx-module#ngxsocketconnect the connection pool will be checked for a valid idle connection to re-use; if a valid idle connection is available, it will be used instead of opening a new connection and performing another three-way handshake.

If you're seeing behavior in your environment that doesn't align what the documentation outlines, it would be helpful to post a minimal, standalone, reproducible example for folks to take a look at.

P.S : This the first time I am posting a message in an opensource community, so please don't mind anything.

Welcome to the open source world! :D