Runtime Error: Algorithm took longer than 10 minutes on a single time loop

I think that error message may be a bit misleading.

IIRC, it's 10 minutes per time step (e.g. each call to `OnData()` has 10 minutes to complete). More expensive functions can be wrapped in `Train()`.

The caveat here is that during a backtest, the resource allocation (30-60 minutes?) is shared across all calls to `Train()` and Scheduled Events per backtest. So if each scheduled event takes ~3 minutes and it fires monthly, you would hit this error after 1-2 years of backtesting. In live trading, this will work just fine.

My suggestion would be to either move some of the logic into `OnData` instead of Scheduled Events, or split the backtest up into multiple backtests (use the ObjectStore if you need to retain states).

Thanks Adam. Just to clarify something I'm confused about:

My suggestion would be to either move some of the logic into `OnData` instead of Scheduled Events,

I'm wondering why would this work? Does the backtester allow OnData to have more time to run than scheduled events? What's confusing me is that you also mentioned: e.g. each call to `OnData()` has 10 minutes to complete,

Also, there's a  chance I'm just hitting an upper resource limit since I'm roughly doing exactly what you described: quarterly calls, taking about 3-4 minutes each, and after 2.5 years it spits this error (even though each quarter is always under 10 minis). Total backtest time is roughly 30 minutes…but I've previously ran backtests over 2 hours without hitting any errors.

Any idea if there's a way to verify that this is truly the issue? I can't seem to find a limit to backtest time in the docs, only the number of nodes/RAM/processing speed etc. 

I'm wondering why would this work? Does the backtester allow OnData to have more time to run than scheduled events? What's confusing me is that you also mentioned: e.g. each call to `OnData()` has 10 minutes to complete,

Right, it's a bit subtle:

• OnData() is allowed 10 minutes per timestep in algorithm time
• Scheduled Events/Train() is allowed a fixed allocation (30-60 minutes) and refills (5-15 minutes/24 hrs) via a leaky bucket in real-time

This means that in backtesting, you basically have 30-60 minutes (depending on server I think) to run all the Scheduled Events/Train per backtest. It's okay if the total backtest time is longer than this.In live trading, algorithm time is the same as real-time so as long as you are under the compute time allocation in the scheduled functions/training it's okay.

--For your specific case (<10 mins each scheduled event), it's a very simple workaround: 

def Initialize(self):
    self.AddEquity('SPY', Resolution.Minute)
    self.flag = False
    
    # Your Scheduled Event
    self.Schedule.On( # scheduling logic, 
                      self.ScheduledFunction )
                      
def OnData(self):
    if self.flag:
        DoSomething()
        self.flag = False
                      
def ScheduledFunction(self):
    self.flag = True

Gotcha, I'll try to put my time-heavy code in OnData then at some point this coming weekend and I'll comment back the results. Hopefully it should work out. Thanks for the explanation Adam!