Hi. I have 20+ different JVM-based µservices deployed in Kubernetes. In Grafana, I display some metrics gathered with Prometheus. I would like to have a correlation between system condition (e.g. number of processed requests, number of finished business tasks) and particular µservices restarts (in a separate panel).
Initially, I wanted to use process_uptime_seconds (a counter) and detect its reset. However, something like:
sum by(application) (delta(process_uptime_seconds[1m])) < 59.8
doesn’t work as expected. I wonder, if there is a simply way to visualize a counter reset (even ignoring that there could be more instances of the same application)?
Btw, I know the annotations in Grafana which probably could be applied for that, however, while it would be nice for a monolith app restarted a few times a week, for µservices with Continuous Delivery, it could make every panel on a dashboard not very readable.