Monitor Kubernetes with Full Context Across Nodes, Pods, and Services
Atatus monitors every cluster, node, pod, container and workload in your Kubernetes environment, with events and container logs alongside, so your team can see what changed and fix it.
A bad rollout, minute by minute.
Pods, replica counts and events on one timeline. Drag it back to see what changed first.
- Replicas desired
- 3
- Replicas available
- 2
- Restarts in window
- 3
- payment-5b7c-h2v4Runningnode-b
- payment-5b7c-x7d2Runningnode-c
- payment-7d9f-k2x8CrashLoopBackOff · 3 restartsnode-b
- NormalScalingReplicaSetdeployment/payment · Scaled up replica set payment-7d9f to 1
- NormalScalingReplicaSetdeployment/payment · Scaled down replica set payment-5b7c to 2
- NormalScheduledpod/payment-7d9f-k2x8 · Successfully assigned shop/payment-7d9f-k2x8 to node-b
- NormalPulledpod/payment-7d9f-k2x8 · Container image "payment:v2.14.0" already present on machine
- NormalStartedpod/payment-7d9f-k2x8 · Started container payment
- WarningBackOffpod/payment-7d9f-k2x8 · Back-off restarting failed container
- NormalScalingReplicaSetdeployment/payment · Scaled up replica set payment-5b7c to 3
- NormalScalingReplicaSetdeployment/payment · Scaled down replica set payment-7d9f to 0
- NormalKillingpod/payment-7d9f-k2x8 · Stopping container payment
- NormalScheduledpod/payment-5b7c-r8n6 · Successfully assigned shop/payment-5b7c-r8n6 to node-a
- NormalStartedpod/payment-5b7c-r8n6 · Started container payment
Drag the timeline · red band: incident open · sample data
Every pod has an owner. Follow it.
Deployments, StatefulSets, DaemonSets, Jobs and volumes, as a tree. Pick anything to see what Atatus tracks for it.
One Helm chart. On every node you choose.
Namespaces and workloads are found on install. Tolerations let the agent onto tainted nodes like GPU and spot pools.
no taint- node-aatatus-agent-8xk2
- node-batatus-agent-c1m7
- node-catatus-agent-t9q4
CriticalAddonsOnly- sys-1no agent pod
sku=gpu:NoSchedule- gpu-1atatus-agent-m6r1
- gpu-2atatus-agent-w2hd
instance-type=spot:NoSchedule- spot-1no agent pod
- spot-2no agent pod
os=windows:NoSchedule- win-1no agent pod
Untolerated taints keep the DaemonSet pod off a node · sample cluster
helm repo add atatus https://atatus.github.io/helm-chartshelm repo updatehelm install atatus -n atatus --create-namespace -f atatus_values.yaml atatus/atatus-agentAfter a changehelm upgrade atatus -n atatus -f atatus_values.yaml atatus/atatus-agent# atatus_values.yaml atatus: license_key: "lic_infra_****" cluster_name: "my-production-cluster" logs_enabled: true daemonset: enabled: true tolerations: # GPU nodes - key: "sku" operator: "Equal" value: "gpu" effect: "NoSchedule"
Filter by namespace, pod or container · Container monitoring
Kubernetes, screen by screen.
Real screens from the Atatus app.
Questions about Kubernetes monitoring
Pricing, deployment, data and the agent. If something is missing, a person will answer it on a demo call.
Kubernetes Monitoring gives you visibility into your clusters, nodes, pods, containers, workloads, events and storage, so you can see what is running, what changed and what is failing.
Node CPU, memory, disk I/O and network; pod and container CPU, memory, network and restarts; available and desired replicas for Deployments, ReplicaSets, StatefulSets and DaemonSets; Job and CronJob status; persistent volume and claim capacity; and Kubernetes events.
Yes. Set a cluster_name for each cluster in atatus_values.yaml and each one appears in Atatus under that name. You can monitor any number of clusters on the same plan.
Install the Atatus agent with its Helm chart. It runs as a DaemonSet with one agent pod per node, finds all namespaces and workloads on install, and collects container logs when logs_enabled is true. Add tolerations if some nodes are tainted, such as GPU or spot pools.
Yes. container_include and container_exclude in atatus_values.yaml match namespaces (kube:), pod names (kube_pod:) and container names (kube_container:), and accept regular expressions. There are separate options for metrics and for logs, and you can drop log lines by pattern. Node and cluster metrics are always collected.
Per host (node): $20 per host per month plus $1 per pod, or $15 per host plus $0.84 per pod with yearly billing. A host counts once it has sent data for 720 hours in a month, and there is no per-user pricing.
Atatus shows each pod's status and restart count, the container's state and reason, such as CrashLoopBackOff, the Kubernetes events around it, such as BackOff, and the container's logs. An alert rule on restart count tells you when a pod keeps restarting.
Alert rules cover pod and container CPU, memory and restart count, node CPU and memory, replicas for Deployments, ReplicaSets, StatefulSets and DaemonSets, Job and CronJob status, and persistent volume capacity. Each pod, node or workload is evaluated on its own and opens its own incident, and you can narrow a rule by cluster, namespace, node or deployment.