numerical instability warning runs away when environment depleted

[eagmon]

If I let the WC run in a small environment for long enough to deplete nutrients, it will sometimes run away with the following output and I need to kill the process. I would like to just end the simulation when this happens. Any recommendations?

Warning: basis matrix is ill-conditioned (cond = 2.94e+13)
Warning: numerical instability (primal simplex, phase II)
Warning: basis matrix is ill-conditioned (cond = 2.57e+13)
Warning: basis matrix is ill-conditioned (cond = 2.94e+13)
Warning: numerical instability (primal simplex, phase II)
Warning: basis matrix is ill-conditioned (cond = 2.57e+13)
Warning: basis matrix is ill-conditioned (cond = 2.94e+13)
Warning: numerical instability (primal simplex, phase II)
Warning: basis matrix is ill-conditioned (cond = 2.57e+13)
Warning: basis matrix is ill-conditioned (cond = 2.94e+13)
...

[jmason42]

I don't really know how the FBA backend works but it seems like you need to do one of three things:

• Somehow detect that there is going to be a problem before there is a problem.
• Change some solver configuration such that it knows to give up and throw an error.
• See if there is some sort of solver callback that you can patch into, and throw if it hits this loop.

I'm guessing this issue is pretty hard to reproduce since you need to run a long series of simulations, but anything you can do to isolate it will make it easier to test and fix.

[tahorst]

This is also discussed in #165. The sims for the paper also had a few failures where the solver was stuck with endless Warning: numerical instability (primal simplex, phase II) messages until the time expired. I think it will be good to end the loop early or switch solvers methods and hopefully get a solution. A general solution still might not work in your case if you have depleted nutrients