DevOps failed. DevOps was supposed to be "devs doing ops" - or at least something close to that. There was never supposed to be a job called "DevOps Engineer". But where did it go wrong and why did everyone need to hire DevOps engineers anyway? More on that tomorrow.

Using AI at your org? See how your teams can improve results and speedup your incident resolution process in just a few minutes! Plus, you can do it using an OSS project. Here's the agenda...

This is what AI for Prometheus alerts looks like. Below is an alert for HighLatency on an HTTP endpoint. If you want access, DM me or leave a comment.

What was wrong with DevOps? It assumed Devs would keep on developing + take on a myriad of new responsibilities & do them as well as dedicated teams. Yes, devs should be *involved* in other areas, but it's hubris to assume the other responsiblities are trivial, not disciplines.

There is no such thing in Kubernetes as a "pod CPU request". Requests and limits are specified per-container - the closest thing to a "pod request" is the sum of all container requests. Now that's changing with a new KEP adding pod-level requests/limits github.com/kubernetes/enh…

What if we made it EASY for devs to do "you build it, you run it" & gave them tools? But didn't ask them to maintain servers, be security experts, and so on. That's the gist of Platform Engineering. It's a healthier approach than "devs do ops". More on this in webinar tomorrow.

What happens in Kubernetes if you do a rolling update but pods are stuck in terminating? Do new pods get created or are the terminating pods still considered running? There's a KEP for that! github.com/kubernetes/enh…

Who is hiring for DevOps leadership positions? (Senior manager or director level.) Asking for a friend!

Coming soon: HolmesGPT integration for Kafka, so you can see faster why Kafka alerts fire and what is causing lag. To request beta access, DM me or leave a comment and I'll get in touch.

Now slightly old, but a pretty interesting discussion on the K8s mailing list that resurfaced today in my inbox

Last year as CEO I made a risky decision. There was hype around AI, but skepticism had never been higher I chose to focus our attention on AI for PrometheusMonitoring alerts. A year later, it paid off. Here is the result and what it means for the future of on-call.

What happens to a PVC when a StatefulSet is deleted? It stays around forever. Now there's a KEP to change that.

What's the most annoying Prometheus alert that is always a false-positive in your experience? (Especially from defaults in kube-prometheus-stack.) I'll go first - TargetDown is almost always PrometheusTargetMisconfigured.

I'll be at KubeCon next week. Will you?

Feedback wanted: which AI investigation of a Prometheus alert 👇 do you find most useful? We're playing with different output formats for HolmesGPT.

Sneak preview - chat with your observability data and Kubernetes cluster

Are you at KubeCon Salt Lake City? watch new video i recorded with Natan Yellin Natan Yellin from Robusta.dev Using Azure AI and HolmesGPT as an AI Assistant for AKS Alerts. Robusta team is at KubeCon! They are close to CNCF store. #holmesGPT AzureTar youtu.be/hZhcD1D-Bgo?si…

Fascinating chat with Natan Yellin on the future of observability: 🤖 How BYOLLM is the way forward 🚀 AI supercharging incident response 📈 HolmesGPT's impact on MTTR All open source, all epic!

In our Virtual Meetup, Natan Yellin from Robusta.dev shares how #VictoriaMetrics is part of the #HolmesGpt architecture, an #AiAgent. youtube.com/watch?v=33z8e6…