How to stop an ib_async script without causing IBGateway issues.

[jackroc97]

The problem that I am having goes something like this:

1. Open IBGateway and authenticate
2. Run my program which uses reqHistoricalData with keepUpToDate=True
3. Program runs fine and data stream is kept up-to-date as expected
4. Stop program using keyboard interrupt (ctrl+c)
5. Run program again
6. Wait for about 1 minute
7. Receive error message: Error 162, reqId 4403: Historical Market Data Service error message:API historical data query cancelled: 4403, contract: Future(conId=603558932, symbol='ES', lastTradeDateOrContractMonth='20250321', multiplier='50', exchange='CME', currency='USD', localSymbol='ESH5', tradingClass='ES')
8. Close IBGateway, then go back to step 1

How can I stop my program at-will without screwing up whatever is going on behind the scenes in the Gateway? Has anyone else had this issue before?

[gnzsnz]

please check this notebook https://nbviewer.org/github/ib-api-reloaded/ib_async/blob/main/notebooks/basics.ipynb at the end it shows how to disconnect.

and the API documentation for disconnect

[mattsta]

That is an odd situation because the gateway should be releasing all data requests/subscriptions/resources when your client disconnects from the gateway.

Feel free send a more complete example, but I don't see how the gateway would be the problem here.

Another potential cause: IBKR does enforce account-wide "pacing limitations" for trying to access too much market data too quickly, so if you start->stop->start->stop the same data feed many times (even across restarts), it's possible their upstream rate limiting is stopping new requests until the server-side service timeout resets your permissions again.

[jackroc97]

Thanks all for the replies. I think this stems somehow from how I am stopping my program. For testing/debugging, I need a way to stop the program ad-hoc, and I have been using ctrl+c to interrupt and stop the program. To test if this was the factor which is causing these problems, I wrote a simple test script:

from ib_async import *

if __name__ == '__main__':
    try:
        ib = IB()
        ib.connect(port=7496)
    
        contract = Future(symbol="ES", lastTradeDateOrContractMonth="202503", exchange="CME")

        hist_data = ib.reqHistoricalData(
            contract, endDateTime="", durationStr="2 D", barSizeSetting="5 secs", 
            whatToShow="TRADES", useRTH=False) # keepUpToDate=True
    
        df = util.df(hist_data)
        print(df)
            
    finally:
        ib.disconnect()

As long as I don't interrupt the program with ctr+c, I can run this simple script over and over again. I believe this rules out the rate-limit hypothesis. Additionally, I don't think I am ever hitting the limits described in the docs.

The problem comes when I run this program but interrupt it before it comes to a stop. Re-running after interrupting the program causes the error I described in my initial post. The only way to reconnect is to close the Gateway, re-open it, and log back in.

I don't really know what to make of this... is there a "safe" way to stop the program while its running?

[August1328]

One of my scripts also uses reqHistoricalData with keepUpToDate=True.

To catch an user interrupt resp. CTRL + C, I always insert an except like this:

try:
    ...
    ib.connect()
   ...
except KeyboardInterrupt:
    ib.disconnect()
    print("Abborted")
    exit()

Maybe this helps... I have never had your problem.

[jackroc97]

Thank you @August1328! Specifically catching the KeyboardInterrupt seems to have solved my problem. Maybe my understanding of how try... finally works is flawed, but I thought finally would execute no matter what.

[August1328]

Glad to hear.

I was just looking through the docs... Shall we add an example to the documentation, how to avoid issues if you use try? For example in the 'Code recipes' page? I am not familiar how to make changes to a repo, but I can give it a try.