Support for max_requests_connection in downstream connections of thrift proxy

[wbpcode]

Support for max_requests_connection in downstream connections of thrift proxy which make the envoy can actively disconnect from the thrift client at the right time.

Consider such a scenario. Assume that the thrift server has opened port 10000 and the IP is 1.2.3.4. But the envoy sidecar of the thrift client initially only had a virtual listener of 0.0.0.0:10000. Then the connection and request of the thrift client will be processed by 0.0.0.0:10000. But after a while, we created a new listener of 1.2.3.4:10000 for the sidecar, and at this time we expect that the new request will be processed by 1.2.3.4:10000. However, because the connection between the thrift client and the sidecar is a keep-alive connection, unless the client actively disconnects, subsequent requests will still be processed by 0.0.0.0:10000.

[rgs1]

It would still not be a clean drain, because there's no way in thrift to drain a connection cleanly afaik. But maybe that's OK for most use-cases?

Happy to review a PR if you want to give this a go @wbpcode.

cc: @fishcakez

[wbpcode]

It would still not be a clean drain, because there's no way in thrift to drain a connection cleanly afaik. But maybe that's OK for most use-cases?

Happy to review a PR if you want to give this a go @wbpcode.

cc: @fishcakez

I will have a try.

[github-actions]

This issue has been automatically marked as stale because it has not had activity in the last 30 days. It will be closed in the next 7 days unless it is tagged "help wanted" or "no stalebot" or other activity occurs. Thank you for your contributions.

[wbpcode]

/no-stale