Monitoring a microservice architecture

Question

I'm designing an architecture that similar to what's described here. The diagram is:

My question is how do you monitor such an architecture where independent pieces compose into a logical unit? It's almost as if we need a monitoring system that checks S3 for .zip files and then polls S3 for the corresponding png files. If after X hours no png files are found then alert.

Is there a tool that does timeseries analysis? Does Prometheus do this?

Answer 1

TL;DR: Prometheus (usually) works by pulling metrics off a server, so I dont see how you could apply it directly onto S3, unless you generate a dynamic page with the number of png's on S3.

in details: The way Prometheus works is by pulling metrics, available as HTTP pages, from servers. Your server will need to publish this special page called /metrics and Prometheus will go there and get its contents.

If you can generate a dynamic public page on S3 that would export the current number of .pngs in your bucket, that this should work. just point Prometheus to it.