SRE or whatever they are calling Ops here, this blog left me with more "Please hire an Ops Principal". That has nothing to do with Elixir.<p>We (Ops type people) have a developed system for gathering metrics, it's Prometheus stack. Instead of integrating with that system, OpsMaru decided it doesn't work and went with their own custom system. You are showing code you were building all CPU metrics that PromQL query easily does and you code assumes 15 second scrapes so if we need higher resolution temporarily, well, sucks to be your customer. Also, if you did Remote Write, you could Remote Write back to a customer if they wanted it. Hell, you could have written a system so we don't need run Prometheus locally since you would scrape everything and send it back to us.<p>Also, you are already running "my company code" so it might be emitting Prometheus metrics so I'm probably running Prometheus already so I can monitor my own code. However, if I wanted to keep an eye on OpsMaru Uplink, I can't because OpsMaru Uplink doesn't appear to have Metric endpoint I can monitor. Maybe your customers are too small to have Ops people but if they did, they are now blind.<p>I want blog article explaining all options tested and what pitfalls you ran into that you settled on this.
LXD seems like an unusual choice when Kubernetes already has cadvisor and strong monitoring integrations. Avoiding extra agents is nice, but does this really scale better than existing solutions like Prometheus and OpenTelemetry?<p>What’s the advantage here beyond keeping things lightweight? Feels like this could hit limitations as complexity grows.
Hey there! Founder of Opsmaru here. I didn't expect the post to get to the first page after it didn't get upvoted when I first posted it. Happy to answer any questions about the product and this post!