grpc c++ client stuck after server restart

[chnex]

I have a grpc c++ client code to connect to a server (external to my code). When the server crashes, my code will detect, restart the server process, and re-create a client object to connect. It works for the first crash. but for the second crash, the client object seems stuck in the first call (I looked at grpc log: the call is invoked, but the message never reaches the server), as well as the later calls. This applies to both async (completion queue receives nothing) and sync calls (stuck). any pointers/helps? Thanks a lot! The server process always use a same unix socket for communication. I delete it every time after detecting the crash and before restarting the server process.

[yas...@google.com]

It might be worth setting the deadline on the RPC. Also, try setting the keepalive timer using the channel args GRPC_ARG_KEEPALIVE_TIME_MS and GRPC_ARG_KEEPALIVE_TIMEOUT_MS