One of the more interesting things I've done for this specific metric is to graph CPU / QPS. This gives you an average CPU cost per query, which can be quite useful to find changes in query patterns.
I have my prometheus scrape both mysqld_exporter and node_exporter to get this.. But since the port numbers are different, the instances don't match. To solve this, I make a new label with relabeling.
For each of the mysql and node jobs:
relabel_configs:
- action: replace
source_labels: [__address__]
target_label: node
regex: ([\w.-]*?):\d+
replacement: $1