“We were essentially able to reduce the cost of that cluster by about 75%. On AWS, DevZero demonstrated they could achieve significantly higher savings than we initially thought possible.”
Mihir Nair
Head of Architecture, Databahn
Multicloud Kubernetes Monitoring
See workload health, network traffic and cost for every cluster on every cloud, with alerts in Slack, Microsoft Teams, Discord, email or webhooks.
Companies who slashed their Kubernetes
spend using DevZero
Monitor Workload Health Across Every Cluster
Use one way to monitor workload health across every cluster and cloud.
- Spot OOM kills and CrashLoops as they happen
- Find workloads hit by CPU throttling
- Track workload restarts across clusters
Track Network Dependencies Across Clusters
See traffic between your workloads and the external services they call.
- Compare Same AZ, Same Region and Cross Region traffic
- Check Internet Egress and External Private traffic
- See traffic and cost for each category
Set Alert Rules For Every Cluster
Choose the problems you want to hear about and get an alert when they happen on any cluster.
- Get alerted when cluster cost passes a USD amount you set
- Catch sudden cost spikes before the bill arrives
- Know when a cluster stops reporting
- Get alerted on application restarts, OOM kills and CPU throttling with no rules to write
Send Alerts To Your Tools
Send alerts to Slack, Teams, Discord, email or any webhook.
- Receive only Critical alerts by default
- Change the minimum severity to get more alerts
- Skip alerts during quiet hours instead of getting them late
See Department Cost Across Clusters And Clouds
Group workloads into departments by namespaces, workload kinds, labels, annotations or name patterns, and see what each costs across all your clusters. See also Kubernetes cost monitoring.
- Span one department across clusters on different clouds
- Count shared workloads in full or split them evenly
- See unassigned cost separately
Profile Every Workload And Node
Profile Workloads
Profile workloads to see CPU and memory percentiles and behavior patterns for every container. Each workload gets a clear classification.
Detect Languages Automatically
See runtime metrics for Java, Node.js, Go, Rust, Python and other runtimes with no code changes. DevZero detects the language for you.
Trace Traffic With eBPF
Trace traffic between workloads with eBPF and map service dependencies without sidecars.
Profile Application Health
Track startup behavior and health signals for every workload, and spot unhealthy apps before your users do.
Monitor API Performance
See request rate, errors and latency for every container with no instrumentation.
Profile Node Groups
Get instance type and size suggestions based on how you really use each node group.
Consolidate Node Groups
Find node groups you can merge or compact, and see the estimated savings before you act.
Attribute Cost To Departments And Workloads
Track Workload Costs
See what every workload costs over any time period, and roll costs up by team or department.
Break Down Network Costs
Split cross-AZ, internet egress and NAT gateway spend, and see which workloads drive it.
Attribute Storage Costs
Price persistent volumes per storage class and attribute them to the workloads that use them.
Apply Your Discounts
Reflect your Savings Plans, Reserved Instances and Committed Use Discounts, so you see what you actually pay.
How DevZero Compares With a Prometheus and Alertmanager Stack
Where each approach is stronger, and what DevZero adds out of the box.
Capability
Prometheus stack
DevZero Monitoring
Cluster cost alerts in USD
Cost attribution by department
No Prometheus, exporters or Alertmanager to run
Restart, OOM kill and CPU throttle alerts with no rules to write
One alert list across clusters, no setup
Alert rules on any custom metric
Native PagerDuty and Opsgenie receivers
What Our Customers Say
Frequently Asked Questions
Questions from platform engineers and SREs evaluating Kubernetes monitoring.
What does DevZero monitor?
It monitors workloads such as Pods, Deployments and StatefulSets, plus nodes, storage and networking. Node and container metrics come from a small agent on each node.
What access does DevZero need in my cluster?
DevZero reads information about your workloads, nodes and storage. The install also grants some write permissions, for example for its own settings. You can review the full list in the install files before you approve it.
What alert types are available?
Cost Threshold (30-day cluster cost reaches a USD amount), Cost Anomaly (the latest daily cost is well above earlier days), Cluster Disconnected (no report for a set number of hours), Application Restart, Out of Memory and CPU Throttle. Rules run every 15 minutes by default.
Where can alerts be delivered?
Slack, Microsoft Teams, Discord, email and generic webhooks. By default only Critical alerts are sent to these channels, and you can change that minimum severity. Alerts that fall inside quiet hours are skipped, not sent later.
Can I set different alert thresholds per cluster?
No. A threshold is set once per alert rule and applies to every cluster connected to your team. Cost alerts are still raised separately for each cluster.
How is this different from Prometheus with Alertmanager?
Prometheus with Alertmanager is the better fit if you need alert rules on any custom metric, PromQL or PagerDuty and Opsgenie receivers, and DevZero does not replace it for those. DevZero adds cluster cost alerts in USD, cost by department and ready-made workload restart, OOM kill and CPU throttle alerts, with no Prometheus, exporters or Alertmanager to run. It does not collect pod logs or application traces.
How much overhead does DevZero add?
These are default requests, not measured usage. The main component requests 200m CPU and 2Gi memory per copy, with 2 copies by default, and each node also runs a small agent.
Does DevZero also monitor GPU utilization?
Yes, on nodes labelled nvidia.com/gpu.present=true it collects NVIDIA GPU metrics. This is monitoring only.
How do departments work across clouds?
A department groups workloads by namespaces, workload kinds, labels, annotations or name patterns, and none of these depend on the cloud provider. A shared workload counts in full for every department that claims it, or is split evenly in the apportioned view.
Install DevZero.Set Up Your First Alert Rule.
Install DevZero in each cluster and set up cost, workload restart, OOM kill and CPU throttle alerts.