tree 31fcf3ca4500f7ebf09029a618148de09acdae0f
parent 188140328dbf1a98159b6531b7fad77df32490fe
author Fabio Ponciroli <ponch78@gmail.com> 1618406319 +0200
committer Fabio Ponciroli <ponch78@gmail.com> 1619002105 +0200

Plugin reports wrong number of failures

The increment of the metric failure should happen
*when* a new failure is happening and NOT when the metric
is read from a callback.

The metrics are now published as soon as a single check is performed,
i.e.: a doCheck() for a *Check class runs.
The metric collector system will just scrape them when needed without
retriggering any calculation of the metrics.

The metrics are now injected in the AbstractHealthCheck, which
takes care of updating them whenever a check runs.

As part of this change the latency is now represented as a Timer, rather
than a Gauge as it was wrongly done before.

Bug: Issue 14385
Change-Id: I5f149f6893289c2874d125a9c04862d53cacd19b