Why the metrics data will be reset?

[zby-universal]

I have some metrics and the label may have many values.

So will the number of the metrics exceed some threshold and all metrics wiil be reset? What is the maximum threshold for the number of monitoring indicators？

[csmarchbanks]

Counter resets typically happen when a service restarts, and are handled by functions such as rate in Prometheus.

[zby-universal]

Actually, I have checked all machines which did't restart and the service just used counter/histogram metric to record data. But from Prometheus service side, the counter metrics data of all machines appear negative value like as follow:
PS: The service is deployed with 20 instance and multi-proccess way, and I use multi-process metrics method to record the data.