Improved documentation for api_concurrency_limit
#540
Conversation
| # HTTP requests (GET, POST, PATCH) used by the users and some NApps like | ||
| # of_lldp, mef_eline. These requests can happen multiple times in a very | ||
| # short period of time. "threadpool" is the default value to take the same | ||
| # values as thread_pool_max_workers["api"]. It can be replaced by an integer. |
There was a problem hiding this comment.
api_concurrency_limit and thread_pool_max_workers["api"] are very related. Thinking in probably mentioning that the difference is that api_concurrency_limit is for the initial connections/HTTP messages for (GET, POST, PATCH) and thread_pool_max_workers["api"] is for the processing of these messages (e.g content) that relates more with database and changes in memory. Is this correct?
There was a problem hiding this comment.
@Alopalao what we need to explain here is that in summary:
We provide two ways to configure the concurrency of the API server and the number of thread pool workers, https://github.com/kytos-ng/kytos/blob/master/kytos/core/api_server.py#L67-L82. In typical cases what users want is to indeed use api_concurrency_limit = "threadpool" same size as the threadpool but we also support cases where the thread pool can be larger but api_concurrency_limit can be lower limiting the capacity of the requests the server can handle.
And the size of the API threadpool also depends on how much the endpoints are being used (which as know are also correlated with how many resources are being created and scalability), here is a comment worth also linking in the docs #489 (comment) of the template.
Closes #504
Summary
Added more documentation to
api_concurrency_limitin the template for the configuration fileLocal Tests
N/A
End-to-End Tests
N/A