I have been encountering a similar issue regarding the TWS application.
My goal is to simulate scenarios where the TWS application experiences a freeze or requires a restart due to memory-related issues, and to ensure a smooth failover to minimize disruptions.
To test this, I manually restart the TWS application after it has been successfully launched and logged in. Once both the TWS application and TWS API are up and running, I initiate a manual restart directly from within the TWS application itself to observe how the reconnection process behaves.
For handling reconnections after such restarts, I have been using the following pieces of code:
void CBOEEngine::Connect() {
std::cout << "Connect to TWS called\n";
dbglogger_ << "Connect to TWS called\n";
dbglogger_ << "TRYING CONNECTING TO HOST_IP " << host_ip << " HOST_PORT " << host_port << " WITH CLIENT ID "
<< clientId << "\n";
//Disconnect is already connected
if (m_pClient && m_pClient->isConnected()) {
DBGLOG_CLASS_FUNC_LINE << "CLIENT ALREADY CONNECTED DISCONNECTING CLIENT & TRYING\n" << DBGLOG_ENDL_FLUSH;
m_pClient->eDisconnect();
}
//std::this_thread::sleep_for(std::chrono::seconds(1));
if (m_pReader) {
DBGLOG_CLASS_FUNC_LINE << "Deleting previous reader instance\n" << DBGLOG_ENDL_FLUSH;
m_pReader->stop();
m_pReader.reset();
}
//std::this_thread::sleep_for(std::chrono::seconds(1));
DBGLOG_CLASS_FUNC_LINE << "After disconnection request connection_result: " << m_pClient->isConnected() << DBGLOG_ENDL_FLUSH;
//! [connect]
bool bRes = m_pClient->eConnect((const char*)host_ip.c_str(), host_port, clientId, m_extraAuth);
//std::this_thread::sleep_for(std::chrono::seconds(1));
bool connection_result = m_pClient->isConnected();
DBGLOG_CLASS_FUNC_LINE << "bRes: " << bRes << " connection_result: " << connection_result << DBGLOG_ENDL_FLUSH;
if (bRes && connection_result) {
printf("Connected to %s:%d clientId:%d\n", m_pClient->host().c_str(), m_pClient->port(), clientId);
//! [ereader]
if(!m_pReader)
dbglogger_ << "Reader class instance is destroyed\n";
if (m_pClient && m_pClient->isConnected()) {
m_pReader = std::unique_ptr<EReader>(new EReader(m_pClient, &m_osSignal));
if (m_pReader) {
m_pReader->start();
} else {
dbglogger_ << "Failed to create EReader instance\n";
exit(1);
}
} else {
dbglogger_ << "Client is not connected. Cannot create EReader\n";
exit(1);
}
//! [ereader]
p_engine_listener_->OnConnect(true);
is_connected = true;
} else {
dbglogger_ << "CONNECTION FAILED " << host_ip << " HOST_PORT " << host_port << " @ CLIENT ID " << clientId << "\n";
dbglogger_ << "RETRY "
<< "\n";
printf("Cannot connect to %s:%d clientId:%d\n", m_pClient->host().c_str(), m_pClient->port(), clientId);
exit(1);
}
return;
// return bRes;
}
In such cases, the reconnection process appears to handle the disconnection more gracefully without triggering segmentation faults or socket closure errors.
