Add support for indicator host up and down

[opcache]

Add support for indicator host up and down

[beorn7]

Could you explain your proposal in more detail?

[opcache]

I use the pushgateway method to push alarm data, and the monitored host node is offline, how do I need an alarm notification？

[opcache]

I use the pushgateway method to push the alarm data, and I cannot see the up indicator from Prometheus, only the indicator of the pushgateway, as follows:

up{instance="127.0.0.1:9091", job="pushgateway", service="monitor"}

In pushgateway mode, other nodes use curl to send monitoring data to pushgateway. When pushgateway confiscates the monitoring data received from node curl, which indicator should be used for alarm notification?

[beorn7]

Fundamentally: The synthetic up metric is a byproduct of the pull-based metric collection. It cannot work with a push-based approach.

You can alert on pushers failing to push by checking on how long ago the last push happened.

[worker24h]

In k8s， when pod restart or teardown，but pushgateway is alived，however prometheus pulled the last valid data by pushgateway. so I hope pushgateway don't cached data in TTL

[Jasmine456]

在k8s中，当pod重启或者teardown时，但是pushgateway是活跃的，而prometheus通过pushgateway拉取了最后的有效数据。所以我希望pushgateway不要在TTL内缓存数据

Can this problem be solved now