I tend to divide "monitoring" into two pieces: alerting ("tell me when XYZ is down or failed") and metrics ("tell me what's going on right now so I can look for

I tend to divide "monitoring" into two pieces: alerting ("tell me when XYZ is down or failed") and metrics ("tell me what's going on right now so I can look for patterns and bottlenecks"). Those are closely tied together, but they're not quite the same thing: there are plenty of products that are really good at one but not good at all at the other. Looks like you're looking mostly for the latter, in which case Prometheus + Grafana would be the free-to-play option I'd reach for first. The new Grafana Alerts looks interesting for covering the "alerting" piece. Cloudwatch is the logical choice for AWS stuff and it does both pretty well (although I don't know whether you need a separate service like PagerDuty to really handle alerts), but it's not free (especially as you do more with it), and some of the other options like a managed provider (Datadog, e.g.) become more attractive as you put more and more money into it.