Details
Description
Currently we only expose a single removal metric ("master/slave_removals") which makes it difficult to distinguish between removal reasons in the alerting.
Currently, a slave can be removed for the following reasons:
- Health checks failed.
- Slave unregistered.
- Slave was replaced by a new slave (on the same endpoint).
In the case of (2), we expect this to be due to maintenance and don't want to be notified as strongly as with health check failures.