Skip to content
Start free

Every cluster, from the node down to the container

The Helm chart runs the collector on every Linux node and reads cluster state, pod and container usage, objects, events, network flows and container logs. Browse pods, nodes and workloads with their metrics, YAML and events, and see usage against requests and limits.

Start freeBook a demoRead the Kubernetes docs
  • Helm chart · DaemonSet
  • Usage vs requests and limits
  • Kubernetes and Docker

How it works

From one Helm install to every pod in the cluster

Everything uses OpenTelemetry's Kubernetes names, such as k8s.pod.name and k8s.deployment.name, so a pod lines up across metrics, logs, traces and profiles.

  1. 01InstallAdd the Helm chart. It runs the collector as a DaemonSet on every Linux node.
  2. 02CollectCluster state, usage, objects, events, network flows and container logs, each with its own switch.
  3. 03BrowseFind any pod, node, namespace or workload, and open its metrics, YAML and events.
  4. 04SizeCompare usage with requests and limits to find idle capacity and pods with no limits.
  5. 05WatchTurn on ready-made monitors for restarts, pending pods, failed jobs and maxed-out autoscalers.

Explorer

Any pod, node or workload, with its metrics, YAML and events

The explorer lists pods, nodes, namespaces, clusters and workloads in a table you can search, filter and group. Open any object for its overview, metrics, YAML and events, over the same time range as the page.

  • Pods show status, restarts, QoS class, and CPU and memory against their limits
  • Nodes show readiness, kubelet version, pod statuses and usage against allocatable
  • The default 15-minute view is the cluster as it is now; wider ranges show what reported in them

Collection

One chart, six streams, a read-only cluster role

The Kubernetes integration bundles six collection streams in one card, each with its own switch. The collector's cluster role can only get, list and watch, and Helm values switch on the extra privileges that container logs and network flows need.

  • Cluster and usage

    Cluster state from the API, and pod, container, node and volume usage from each kubelet.

  • Objects and events

    Full YAML for pods, workloads, services, nodes and namespaces, and events such as OOM kills.

  • Network flows

    Throughput, retransmits and round-trip time between pods, with eBPF or socket polling.

  • Container logs

    Every pod's stdout and stderr, with its pod, namespace and owning workload.

Containers

Containers and images across Kubernetes and Docker

The container list covers the containers in every cluster and on Docker hosts, with status, restarts and CPU and memory for each. Images roll up by tag, so you can see where each version runs.

  • Docker hosts report through the Docker integration, over the Docker socket
  • Utilization compares each pod's CPU and memory with its requests and limits
  • Ready-made monitors cover restarts, pending pods, failed jobs and autoscalers at their maximum

Better together

The cluster, next to the services that run on it

Pods and nodes share a workspace with your traces, logs and cloud accounts, so a slow service leads to the pod, the node and the cloud cluster behind it.

What it collects

Six streams, each with its own switch

Every stream is on by default and renders as an ordinary OpenTelemetry receiver in the collector's configuration.

StreamWhat you getSourceChart setting
Cluster metricsPod, workload, node and namespace stateKubernetes APINone
Pod and container usageCPU, memory, filesystem and network against requests and limitsEach node's kubeletNone
Object inventoryFull YAML for pods, workloads, services, nodes and namespacesKubernetes API, every 60 secondsNone
Cluster eventsScheduling, OOM kills, probe failures and image pullsKubernetes APINone
Network flowsThroughput, retransmits and round-trip time per connectioneBPF, or socket pollingcnm.enabled
Container logsstdout and stderr with pod, namespace and workloadEach node's container runtimecontainerLogs.enabled

The chart runs on Linux nodes, with a Kubernetes or OpenShift distribution setting. Network flows resolve processes and Service addresses when cnm.enabled is set; container logs need containerLogs.enabled, because the runtime writes them for root only.

FAQ

Questions platform teams ask first

Talk to an engineer
Does the Helm chart run on Windows nodes?
No. The chart runs the collector as a DaemonSet on Linux nodes only.
What access does the collector need in my cluster?
Read-only access to the Kubernetes API: get, list and watch on the objects it reports. When you turn on managed upgrades or uninstall, a separate role lets it update or remove its own DaemonSet, and nothing else.
Can I change workloads from Tracix?
No. The Kubernetes views are read-only. You can read YAML, metrics and events, but not edit, scale or restart anything.
How current is what I see?
The default 15-minute view is the cluster right now: objects are read every minute and dropped after five minutes without a sighting. Wider or custom ranges show what reported inside the range, so a deleted pod stays in a historical view.

See your cluster in minutes

Install the Helm chart with an enrollment token, and pods, nodes and workloads appear as the first objects are read.

Start freeRead the install guide