disbale client keepalive (#92)

Signed-off-by: xufei <[email protected]>

include/pingcap/kv/internal/conn.h

@@ -13,22 +13,9 @@ struct KvConnClient
 
     KvConnClient(std::string addr, const ClusterConfig & config)
     {
-        // set max size that grpc client can recieve to max value.
         grpc::ChannelArguments ch_args;
+        // set max size that grpc client can recieve to max value.
         ch_args.SetMaxReceiveMessageSize(-1);
-        ch_args.SetInt(GRPC_ARG_KEEPALIVE_TIME_MS, 10 * 1000);
-        /// sine we set GRPC_ARG_KEEPALIVE_PERMIT_WITHOUT_CALLS to 1, client channel will keep sending keepalive pings when
-        /// there is no outstanding streams, and in the implementation of grpc-core, for client channnels without activing
-        /// polling threads(which in my understanding, client channel without any outstanding streams will eventually get
-        /// into this situation), it uses backup poller to poll the message(including the ping ack), and the backup polls
-        /// are run in the timer threads and the default poll interval is 5000ms. So we need to set the keepalive timeout
-        /// as (5000 + x)ms, where x ms is the shortest timeout threshold in the worst case. Refer to
-        /// https://github.com/pingcap/tiflash/issues/4192#issuecomment-1073401848
-        /// for more details
-        // todo we can adjust the timeout threshold once we change the default value of backup poll interval
-        ch_args.SetInt(GRPC_ARG_KEEPALIVE_TIMEOUT_MS, 8 * 1000);
-        ch_args.SetInt(GRPC_ARG_KEEPALIVE_PERMIT_WITHOUT_CALLS, 1);
-        ch_args.SetInt(GRPC_ARG_HTTP2_MAX_PINGS_WITHOUT_DATA, 0);
         ch_args.SetInt(GRPC_ARG_MIN_RECONNECT_BACKOFF_MS, 1 * 1000);
         ch_args.SetInt(GRPC_ARG_MAX_RECONNECT_BACKOFF_MS, 3 * 1000);
         ch_args.SetInt(GRPC_ARG_INITIAL_RECONNECT_BACKOFF_MS, 1 * 1000);