|
|
@ -1161,6 +1161,22 @@ static void listener_destroy_done(void* s, grpc_error* error) { |
|
|
|
gpr_mu_unlock(&server->mu_global); |
|
|
|
gpr_mu_unlock(&server->mu_global); |
|
|
|
} |
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
|
|
|
- Kills all pending requests-for-incoming-RPC-calls (i.e the requests made via |
|
|
|
|
|
|
|
grpc_server_request_call and grpc_server_request_registered call will now be |
|
|
|
|
|
|
|
cancelled). See 'kill_pending_work_locked()' |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
- Shuts down the listeners (i.e the server will no longer listen on the port |
|
|
|
|
|
|
|
for new incoming channels). |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
- Iterates through all channels on the server and sends shutdown msg (see |
|
|
|
|
|
|
|
'channel_broadcaster_shutdown()' for details) to the clients via the |
|
|
|
|
|
|
|
transport layer. The transport layer then guarantees the following: |
|
|
|
|
|
|
|
-- Sends shutdown to the client (for eg: HTTP2 transport sends GOAWAY) |
|
|
|
|
|
|
|
-- If the server has outstanding calls that are in the process, the |
|
|
|
|
|
|
|
connection is NOT closed until the server is done with all those calls |
|
|
|
|
|
|
|
-- Once, there are no more calls in progress, the channel is closed |
|
|
|
|
|
|
|
*/ |
|
|
|
void grpc_server_shutdown_and_notify(grpc_server* server, |
|
|
|
void grpc_server_shutdown_and_notify(grpc_server* server, |
|
|
|
grpc_completion_queue* cq, void* tag) { |
|
|
|
grpc_completion_queue* cq, void* tag) { |
|
|
|
listener* l; |
|
|
|
listener* l; |
|
|
|