GRPC_CALL_ERROR_TOO_MANY_OPERATIONS Error on Concurrent Write Operations

[szanna]

Hello,

I am facing an issue with a service where multiple clients are connecting and the server is sending data to all of them. Here is a brief outline of my service definition:

service StreamService {
  rpc Connect(ConnectRequest) returns (stream Data);
}

In my code, I have a handler which looks like this:

auto [channel1, channel2] = observers.addObserver(co_await asio::this_coro::executor);
auto [a, b] = co_await (myWriter(writer, channel1) && myWriter(writer, channel2));

The same myWriter function is used twice because each receives different types of data from different asio::channel, but they are going to the same stream Response. The

Here is the myWriter function:

const auto [data] = co_await channel.async_receive(...);
bool ok = co_await agrpc::write(writer, data);

However, I am getting a GRPC_CALL_ERROR_TOO_MANY_OPERATIONS error.

As I understand it, for a single client connection, gRPC can't send another message until the previous one is finished. In that case, does asio-grpc, specifically agrpc::write, handle this issue on its own? If so, where else should I look for a problem if I am incorrectly using asio-grpc?

Additionally, I am curious if there is a function available that checks whether a stream is in progress. This could be helpful in understanding how this error situation occurred.

I appreciate any help or guidance you can provide. Thank you for your time.

Best regards,

[Tradias]

Hi,

thanks for the detailed error report. You are correct, gRPC does not support sending another message until the previous one is finished and neither does asio-grpc. agrpc::write is just a thin wrapper around grpc::ServerAsyncWriter.Write(). The same limitations apply. They are also documented https://tradias.github.io/asio-grpc/structagrpc_1_1detail_1_1_write_fn.html#a87ab5914b5bac633e1cdb3798ba85e6e (which is essentially just a copy of the official gRPC documentation).
You have to synchronize the writes yourself, either by having both observers write to the same channel or by creating a third channel that collects the other two.

The is no function to check whether a stream is in progress. But since you are using coroutines it should be easy to write your own.

[Tradias]

Was my information helpful? Did you manage to resolve your issue?

[Tradias]

Closing, please do not hesitate to create a new issue if you face any more problems.