reverse_proxy: active health check should allow configurable fails and passes
ottenhoff opened this issue · comments
Use Case: legacy Java app will sometimes fail a single active health check (maybe because of GC), but we don't want to mark the upstream as down until at least 3 checks fail in a row. Then we don't want to mark the upstream as healthy until as least 2 passes in a row succeed to make sure that we aren't marking a dying server as healthy.
Possible terminology:
- fails / passes
- fall / rise
- healthy_threshold / unhealthy_threshold
Sounds like a reasonable request. I'm quite busy at the moment but anyone could work on this!