kube-prometheus VS descheduler

The last one is mostly an observability stack with Prometheus, Metric server, and Prometheus adapter to have excellent insights into what is happening on the cluster. You can reuse the same stack for autoscaling by repurposing all the data collected for monitoring.

kustomize Kube Prometheus

On the other hand, the Kube-prometheus project provides documentation and scripts to operate end-to-end Kubernetes cluster monitoring using the Prometheus Operator, making easier the process of monitoring the Kubernetes cluster.

For our load testing we’ve deployed Temporal on Kubernetes, and we’re using MySQL for the persistence backend. The MySQL instance has 4 CPU cores and 32GB RAM, and each Temporal service (Frontend, History, Matching, and Worker) has 2 pods, with requests for 1 CPU core and 1GB RAM as a starting point. We’re not setting CPU limits for our pods—see our upcoming Temporal on Kubernetes post for more details on why. For monitoring we’ll use Prometheus and Grafana, installed via the kube-prometheus stack, giving us some useful Kubernetes metrics.

The 2 most common approaches I have seen are kube-prometheus-stack and kube-prometheus..

- I'm missing a lot of the Grafana dashboards that are provisioned during the deployment, not sure why as it has worked before, and wanted to add them after install... I believe it's different ConfigMaps like the one in kube-prometheus but I was wondering if there's a way to force provisioning them all again at once (multiple k8s, node_exporter, vm, etc)?

Check out the kube Prometheus project -- https://github.com/prometheus-operator/kube-prometheus It's a bit heavy, but the included recording rules and dashboards give you a great start at understanding your cluster.

The actual link to the prometheus/grafana bundle: https://github.com/prometheus-operator/kube-prometheus

Here’s a list of what’s installed: https://github.com/prometheus-operator/kube-prometheus/tree/main/manifests

Problem : Uneven Pod Spreading Developers observed uneven pod distribution, especially during traffic spikes, despite configuring PodDisruptionBudget(PDB) and PodTopologySpreadConstraints(PTSC) with whenUnsatisfiable set to ScheduleAnyway. Employing the descheduler, which periodically checks the cluster and evicts pods from overcrowded nodes, forcing them to reschedule onto less crowded ones. Simple, effective, and automated.

maybe descheduler can help? https://github.com/kubernetes-sigs/descheduler

Surprised this wasn't suggested yet, you can also use a software like the k8s Descheduler that executes periodically to rebalance your workloads across the existing nodes.

Here an example of coordination api in Go https://github.com/kubernetes-sigs/descheduler/commit/3cbae5e72ba53447a609e6001755ff395e6eeceb https://github.com/kubernetes-sigs/descheduler/commit/0a52af9ab82a52fd8c864a81f4033736f11aab34

This is something a (now former) colleague of mine pointed out: that the kubernetes descheduler can enforce a maximum lifetime[0] that sort of forces continual reboots. So if your system cannot tolerate running for a long time continously, this is one method to gracefully restart long running pods.

[0]: https://github.com/kubernetes-sigs/descheduler#podlifetime

no it does not... thats why tools like https://github.com/kubernetes-sigs/descheduler exist..