Best practices for creating / retaining / reusing clients?

[AttilaTheFun]

Hi! I was wondering what the best practices are with regards to when I should create new clients vs. recycle existing ones? Also I'm assuming all clients should generally share one EventLoopGroup.

Should I create one client per GRPC service and reuse the same client for subsequent or parallel requests to that service?

Or should I create one client per request, and discard them once the request succeeds / fails? (Ignoring the streaming case where it might be longer lived.)

And if I do create a client per request, do I need to retain it myself until the request completes? Or will it remain in memory long enough to call the completion handler?

Thanks!

[glbrntt]

Hi! I was wondering what the best practices are with regards to when I should create new clients vs. recycle existing ones? Also I'm assuming all clients should generally share one EventLoopGroup.

Clients are reasonably expensive to create and are intended to be long-lived objects so create as few as possible. If a client isn't being used (i.e. there are no RPCs in progress) its connection will be torn down after a period of inactivity. When you next use that client a new connection will automatically be established. Clients can share an EventLoopGroup and in the vast majority of cases that EventLoopGroup only needs a single EventLoop.

Should I create one client per GRPC service and reuse the same client for subsequent or parallel requests to that service?

Yes

[AttilaTheFun]

Awesome, thanks!