Prometheus v3 (introduced in 4.19) defaults to UTF-8 for metrics validation and escaping.
However, the monitoring stack is not yet fully ready for this change: the Prometheus Operator does not fully support UTF-8.

As a result, the current behavior is inconsistent. For example, a UTF-8 target can be scraped successfully, but adding recording rules or alerts for it may not be possible. See https://issues.redhat.com/browse/MON-4280.

Even once UTF-8 support is complete, we may want it disabled by default, requiring users to explicitly enable it per target. Otherwise, a target silently switching from `foo_bar` to `foo.bar` could be confusing.

4.x updates should avoid introducing such breaking changes.

We should default to the legacy validation scheme for scraping.

Note that promtool e.g. may still use UTF-8 for validation.

    

    1.
    2.
    3.