gnosis / dex-services / 5265 / 3

31%

master: 31%

push

travis-pro

Record a metric for when the solver ran out of time (#1462)

This will help us to measure one of our goals ("Batches with non trivial solutions in which Standard Solver (using Gurobi), Best Ring Solver (using SCIP) and Open Solver finish without interruption before time limit").

At the moment this could be inferred by the optimality gap (close to 0 meaning solved to optimality, but it's not entirely clear what the right threshold is, therefore I think it's easy enough to add this metric as an explicit counter.

Here is an example instance that was interrupted because it ran out of time: https://gnosis-dev-dfusion.s3.amazonaws.com/data/rinkeby_dev/standard-solver/results/2020-10-04/instance_5339384_2020-10-04T12%3A45%3A30.550230225%2B00%3A00/06_solution_int_valid.json

### Test Plan

Run:

```
cargo run --bin driver -- --node-url https://staging-openethereum.mainnet.gnosisdev.com  --rpc-timeout 60000 --orderbook-file target/orderbook_mainnet.bin --private-key c87509a1cc0241e5e4a9ec0a0f44dc0d3 --solver-type StandardSolver  --latest-solution-submit-time 70 --scheduler SYSTEM
```

to have the SCIP solver run with minimal time (causing an interrupt). Then check `http://localhost:9586/metrics`
Observe

`dfusion_solver_interrupted 1`

13096 of 53841 branches covered (24.32%)

Branch coverage included in aggregate %.

3476 of 5436 relevant lines covered (63.94%)

43546.17 hits per line