Rule etcdInsufficientMembers not properly working in v3.5?

[JoseL7]

Hi there,
I found that in a three members etcd stacked cluster (k8s), with one master node down the alert etcdInsufficientMembers is triggered with only one member down, in the rule of v3.5:
expr: sum(up{job=~".*etcd.*"} == bool 1) without (instance) < ((count(up{job=~".*etcd.*"}) without (instance) + 1) / 2)

This part "sum(up{job=~".etcd."} == bool 1) without (instance)" returns a series of 3 items:

{container="nginx", endpoint="http-metrics", job="kube-etcd", namespace="kube-system", pod="prometheus-scrape-services-hrcz6", service="prometheus-etcd-metrics"}	0
{container="nginx", endpoint="http-metrics", job="kube-etcd", namespace="kube-system", pod="prometheus-scrape-services-28qwp", service="prometheus-etcd-metrics"}	1
{container="nginx", endpoint="http-metrics", job="kube-etcd", namespace="kube-system", pod="prometheus-scrape-services-dkfl8", service="prometheus-etcd-metrics"}	1

and "(count(up{job=~".etcd."}) without (instance) + 1) / 2" retunrs:

{container="nginx", endpoint="http-metrics", job="kube-etcd", namespace="kube-system", pod="prometheus-scrape-services-hrcz6", service="caas-prometheus-etcd-metrics"}	1
{container="nginx", endpoint="http-metrics", job="kube-etcd", namespace="kube-system", pod="prometheus-scrape-services-28qwp", service="caas-prometheus-etcd-metrics"}	1
{container="nginx", endpoint="http-metrics", job="kube-etcd", namespace="kube-system", pod="prometheus-scrape-services-dkfl8", service="caas-prometheus-etcd-metrics"}	1

First series (0<1) returns TRUE and alert is triggered.

In previous version v3.4, works properly, alert is not triggered with one master node down:
expr: sum(up{job=~".*etcd.*"} == bool 1) by (job) < ((count(up{job=~".*etcd.*"}) by (job) + 1) / 2)

Has anyone else also observed this?