Control over connection pooling?

[bennofs]

Right now, it seems that connection pooling is interwired with the http client itself. Does that have to be the case? Would it not be possible to implement pooling with a custom connector instead?

I have a program that generates a lot of connections, even though it only ever connects to the same server. I need to investigate a little bit more, but I'm assuming right now that for some reason, it is losing the race with the pool checkout quite often.

[mraerino]

@seanmonstar is there any work planned on this?

looks like the connection pool sits above the Connector instead of inside it, so you can't easily replace it right now.

which solution would you rather accept a PR for?

• move connection pooling into connector
• allow setting custom connection pool impl in Builder via some trait

[davidbarsky]

@mraerino I can't comment if work on this is planned, but the the lower-level hyper::client::conn module would allow you build a high-level client with arbitrary pooling. If you chose to go down this path, I'd suggest looking at the internals of high-level Client.

[mraerino]

thanks for the suggestion!
a colleague suggested the same and i already implemented it that way with deadpool.

The boilerplate is actually pretty minimal, great stuff!