We have an issue with duplicate sequence numbers - occasionally an outgoing heartbeat message has the same sequence number as another outgoing message.
Apparently the cause is that we periodically call call_hb_service from one thread (our fix8 message loop thread), but send messages on other threads.
You recommended to protect these operations with a common lock.
Currently, we are using session-scoped locks around calls to send, i.e. sending messages to different sessions in parallel is allowed and works fine.
However, call_hb_service is called for a bundle of sessions (i.e. all that are hadled by the same SessionManager), so we would have to replace the session-scoped lock with a global lock (or at least a SessionManager-scoped lock). I'm worried that this would have a significant performance impact (in production, we are connected to 30 sessions or more).
Is there a solution to this dilemma (other than replacing call_hb_service with per-sessions calls to heartbeat_service)?