Added client keepalive arguments to the grpc-proxy #17366
Conversation
…m arguments to the grpc-proxy Signed-off-by: shihuixing <shihuixing@jd.com>
|
Hi @halleyshx. Thanks for your PR. I'm waiting for a etcd-io member to verify that this patch is reasonable to test. If it is, they should reply with Once the patch is verified, the new status will be reflected by the I understand the commands that are listed here. Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. |
|
please recheck,thanks @ahrtr |
|
/ok-to-test |
Signed-off-by: shihuixing <shihuixing@jd.com>
|
If you backport the PR to 3.5 (I have no objection on it), then add a changelog for 3.5 after you raise the backporting PR. thx |
Signed-off-by: shihuixing <shihuixing@jd.com>
OK,I updated the changelog-3.5. Please recheck. thx |
PLease read my previous comment . Pls do not update changelog in this PR. Instead, update the changelog after raising the backporting (to 3.5) PR |
Signed-off-by: shihuixing <shihuixing@jd.com>
Ok. removed. But How should I raise the backporting (to 3.5) PR? |
Make the same change for https://github.com/etcd-io/etcd/tree/release-3.5. |
Happy new Year! |
Update https://github.com/etcd-io/etcd/blob/main/CHANGELOG/CHANGELOG-3.5.md#v3513-tbd |
thx. Happy new year! |
Sorry! Please check #17447. |
Pls create a new PR. |
OK, Pls check #17448 . thx! |
Added dial-keepalive-time、dial-keepalive-timeout、permit-without-stream arguments to the grpc-proxy for KeepAlive ClientParameters.
Currently cluster's request link: clients(A、B、C) --> grpc-proxy(start by c1.etcd.com) --> etcd-cluster-domain(c1.etcd.com) --> loadbalancer(11.11.11.11) --> etcd endpoints(e1、e2、e3).
When loadbalancer(11.11.11.11) died, new loadbalancer(11.11.11.12) to replaced, grpc-proxy can't provide service to his clients(A、B、C) for 15min.（why 15min？because gRPC balancergroup's balancerCache is default 15min in gPRC-1.38.0
var DefaultSubBalancerCloseTimeout = 15 * time.Minute), it's too late.So we need to actively probe for loadbalancer of etcd-cluster-domain survival and reconnect.
Currently added client keepalive arguments to the grpc-proxy.
This PR introduces three options dial-keepalive-time、dial-keepalive-timeout、permit-without-stream for grpc-proxy：
dial-keepalive-time allows to specify client keepalive interval for grpc-proxy itself (default 0-disable).
dial-keepalive-timeout allows to specify client keepalive timeout for grpc-proxy itself (default 20s).
permit-without-stream allows to specify client sends keepalive pings even with no active RPCs for grpc-proxy itself (default false).
When I start grpc-proxy with
--dial-keepalive-time '20s', It worked.