You can now get alerted when pods get stuck, a namespace nears its quota or a node runs short of capacity, with the same Atatus Alerting rules you already use for Kubernetes CPU and memory. New metric types cover pod status, cluster nodes, namespaces, warning events and PVC volumes.
Scope any of these rules with the Filter By field: type a namespace, pod or other resource name and pick from an autocomplete list of what is running in your cluster, so a rule watches only the workloads a team owns.
What's new in Kubernetes alert rules
- Pod status: Alert on Pod Not Ready, Pod Failed and Pod Pending counts, and on Replicas Unavailable for deployments, so stuck rollouts surface quickly.
- Node and cluster capacity: Track node CPU, memory, disk and pod allocation as a percentage, nodes that are not ready or report problem conditions, and how much of the cluster's CPU and memory is already requested.
- Namespace quotas: Get warned before a namespace hits its resource quota, with CPU and memory requested or used, and pods used, as a percentage of quota.
- Warning events: Count Kubernetes warning events such as BackOff, Evicted and OOMKilling, and alert when they climb.
- PVC volumes: Alert when a persistent volume claim runs low on free disk space, in bytes or as a percentage.
- Scoped to the right workloads: Filter By narrows a rule by namespace, pod, node, cluster or workload, so on-call engineers are not paged for unrelated services. See Kubernetes monitoring for the data Atatus collects.