Awesome. That clarifies my questions. Thanks! I'm trying to get a sense as 
to what I should set connection limit on a heroku 1x dyno. I ran ulimit -a 
within the dyno using heroku run and saw a value of 10000 for the max # of 
file descriptors. 100 does seem very conservative, as mentioned in the 
documentation. I don't want to set this value to something unsafe, but I 
would like to maximize the number of open connections per dyno. Do you have 
any advice in what I should set that value? I don't anticipate needing to 
support > 100 connections at once super soon, but would like to plan ahead.

On Monday, October 5, 2020 at 9:23:25 PM UTC-4 [email protected] wrote:

> The connection limit dictates how many individual tcp connections waitress 
> will handle at a time, and while those are alive (until client hangs up or 
> idle channel timeout) no other connections will be made. The backlog is a 
> signal to the OS to not outright reject connections even if waitress is not 
> willing to handle them yet.
>
> From the list of connections, waitress will handle requests based on the 
> number of threads.
>
> On Oct 5, 2020, at 20:06, Cooper Baird <[email protected]> wrote:
>
> I am starting to use Waitress, and I am trying to understand how channels 
> and the backlog work, so forgive me for my ignorance if I'm not 
> understanding this correctly. Let's say, hypothetically, that I am using 
> all of the default settings (so 100 connection limit, 1024 backlog 
> capacity, 4 threads, etc.). Let's say 100 users, all using HTTP/1.1 
> clients, go to the site at once and begin browsing. Does this mean that any 
> additional users (past the 100) that try to browse the site will hit an 
> error or have a connection timeout since the 100 users fill up the channel 
> capacity of 100 (and being HTTP/1.1 clients, all their requests will be 
> served over the same channel, keeping it open)? If this is the case, then 
> does that mean anyone past those initial 100 users will have to wait some 
> time between 30s (cleanup interval) and 120s (channel timeout) to be able 
> to browse? Or is this where the backlog comes in and channels can be reused 
> somehow between users/clients? I apologize if that didn't all make sense. I 
> can clarify anything that was unclear in my thought process/questioning. 
>
> -- 
> You received this message because you are subscribed to the Google Groups 
> "pylons-discuss" group.
> To unsubscribe from this group and stop receiving emails from it, send an 
> email to [email protected].
> To view this discussion on the web visit 
> https://groups.google.com/d/msgid/pylons-discuss/b9870007-07ea-4e25-bbd0-266e6d05bac2n%40googlegroups.com
>  
> <https://groups.google.com/d/msgid/pylons-discuss/b9870007-07ea-4e25-bbd0-266e6d05bac2n%40googlegroups.com?utm_medium=email&utm_source=footer>
> .
>
>
>

-- 
You received this message because you are subscribed to the Google Groups 
"pylons-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/pylons-discuss/2b78006d-50f1-4acb-a7e7-4d722e372244n%40googlegroups.com.

Reply via email to