Crisper examples of runtime metrics.
jswrenn opened this issue · 0 comments
jswrenn commented
For each task metric, it's fairly easy to write a crisp, self-contained example that reliably induces a change in a metric. For runtime metrics, it's currently not so easy to do this, because:
- runtime metrics are buffered
- some runtime metrics are dependent on scheduling pathologies that are finicky to induce
We could resolve the first obstacle to provide some mechanism to flush metrics on-demand. For the second obstacle, I'm not sure there's much we can do.