What to collect?
We started with already provided Spring Boot metrics. Since we wanted to transform the input a little bit (compute diffs between snapshots, count summary hits for various status codes etc.) we made simple utility tool that reads metrics and feeds the Elasticsearch. And we also implemented mail notifications in case of errors. Maybe we could use Logstash, however it does look straightforward.
We simply used official Docker images. Once it is running, you just need to setup Kibana indices and you can play with your visualizations. In most cases you want to use Date Histogram but you can also get interesting statistics also from Terms aggregation.
The result can look like following: