Metoro Kubernetes Profiling

Metoro Kubernetes Profiling is a continuous Kubernetes profiler built around kernel-level eBPF sampling. It fits platform and application teams that need CPU and memory visibility without changing application builds, including workloads supplied by vendors. Its strongest advantage is the combination of low intended overhead and Kubernetes-aware context; the main buying question is whether per-node and excess-data charges fit your fleet.

Overview

Metoro profiles workloads continuously with an eBPF sampler running in the kernel. Instrumentation requires no SDK, application-code change, language agent, restart, or rebuild, so it can cover both software your team owns and third-party containers you cannot recompile. Sampling runs at 97Hz per CPU core and is designed for less than 0.5% CPU overhead per node, making the approach suitable for ongoing observation rather than only occasional diagnostic sessions.

CPU and memory profiles are connected to Metoro APM traces, logs, metrics, pod state, and deployment context. That linkage gives an investigation more operational framing than a standalone profile and can help relate a hotspot to a deployment or service. Hobby and Scale retain data for 28 days.

Key features

Kubernetes-focused flame graphs

Interactive flame graphs can be aggregated by service, pod, or container, then filtered by cluster, namespace, deployment, or Kubernetes label. These dimensions are useful for narrowing a fleet-wide problem to a particular workload or release. Continuous alerts add a way to surface profiling conditions instead of relying solely on manual inspection.

Runtime and workload coverage

Metoro lists Go, Python, C++, C, Rust, Node.js, and Java support. Other runtimes can still expose host-level frames, although the listed runtime support is the clearer fit when language-level attribution matters. Because profiling is eBPF-based, vendor and other third-party workloads are within scope even when their source and build pipeline are outside your control.

Operations and integrations

Installation uses Helm, and deployment choices are Metoro Cloud, bring-your-own-cloud (BYOC), or on-premises. Metoro Cloud integrates with Slack, PagerDuty, Datadog, Grafana, webhooks, REST APIs, and OpenTelemetry. Its GitHub App can create pull requests with suggested fixes and GitHub issues from investigations, which is most useful to teams that already route remediation through GitHub.

Metoro Cloud states SOC 2 Type II certification, TLS 1.3 in transit, AES-256 at rest, SSO/SAML 2.0, role-based access control, and immutable audit logs. Scale support includes Slack access to Metoro engineers; Enterprise adds 24/7 white-glove support and onboarding.

Pricing

Hobby

Hobby is free forever at no cost. It covers one cluster, one user, and two nodes with 28-day retention. This is a credible way for an individual or very small Kubernetes environment to evaluate continuous profiling, but the node, cluster, and seat caps make it unsuitable for a larger shared platform.

Scale

Scale costs $20 a month, billed at $20 USD / node / month, with an additional $0.20 / GB on excess over 100GB / node. It provides unlimited clusters, users, and nodes plus 28-day retention. The unlimited resource counts remove Hobby's structural caps, but node-based billing means cost grows with fleet size and the excess-data charge matters for high-volume profiling.

Enterprise

Enterprise has custom pricing and is billed by contract, with BYOC or on-premises deployment, unlimited clusters, users, and nodes, custom retention, and custom SLAs. It is the appropriate tier for organizations that require negotiated operational terms or retention beyond the standard 28 days. A 14-day trial is offered for evaluating the paid service.

Platforms

Metoro is offered through web and API access, with Linux and self-hosted deployment options. It supports EKS, GKE, AKS, OpenShift, Rancher, k3s, and self-managed CNCF-conformant Kubernetes distributions. eBPF requires Linux kernel 4.18 or newer, so clusters on older kernels are outside the stated requirement.

Who it's for

Choose Metoro when Kubernetes is the operating boundary and you want continuous CPU and memory profiles correlated with traces, logs, metrics, and deployment state. It is particularly well suited to platform teams, organizations running vendor containers, and engineering groups that cannot justify instrumenting every service. Hobby fits a two-node experiment; Scale fits teams needing unlimited clusters and users; Enterprise fits controlled BYOC or on-premises environments with custom retention or SLAs.

Look elsewhere if your priority is a broad, host-oriented monitoring service rather than Kubernetes profiling, if your environment cannot meet the Linux kernel requirement, or if per-node and excess-data billing is difficult to forecast.

Pros and cons

Pros

• No code instrumentation: SDKs, agents, rebuilds, restarts, and application changes are not required.
• Low intended overhead: 97Hz sampling is designed for less than 0.5% CPU overhead per node.
• Useful Kubernetes dimensions: Service, pod, container, cluster, namespace, deployment, and label views make narrowing investigations practical.
• Third-party coverage: eBPF can profile containers and vendor workloads you do not build.
• Deployment and security choice: Cloud, BYOC, and on-premises options are paired with stated enterprise controls.

Cons

• Small free tier: Hobby is limited to one cluster, one user, and two nodes.
• Usage-sensitive Scale pricing: $20 per node per month is supplemented by $0.20 per excess GB over 100GB per node.
• Kernel dependency: eBPF requires Linux kernel 4.18 or newer.
• Runtime caveat: Runtimes outside the named list are represented at host level rather than with the listed language coverage.

Alternatives

For a wider comparison, see the Continuous Profiling Tools and eBPF Observability Tools lists.

• Guance Container Monitoring is a freemium, multi-platform option whose free plan covers up to 10 hosts, 10,000 metric series per day, 10 million log records per day, and seven-day rolling retention; choose it when general container metrics and logs matter more than Metoro's profiling context.
• Middleware offers a free-forever plan with up to 100GB of data, 1,000 RUM sessions, and 20,000 synthetic checks; it is a better fit for teams prioritizing broader application experience and testing coverage.
• Grafana Cloud Application Observability has an always-free tier with 2,232 host hours and a Pro plan priced at $0.025 per host hour plus a $19 monthly platform fee; choose it when Grafana-centric application observability and host-hour pricing suit you.
• Coroot provides a self-hosted Apache 2.0 Community Edition and a Standard plan billed per monitored CPU; it suits teams seeking open-source, self-hosted observability.
• LMbench is free GPL-2.0 benchmark software for Linux and Windows, intended for benchmarking rather than continuous Kubernetes profiling.
• OneUptime is an open-source Apache 2.0 platform that can be self-hosted; pick it when inspectable source and broader incident tooling are the priority.
• Parca is free, Apache 2-licensed profiling software; it is the natural alternative for teams wanting an open-source profiling stack.
• Pyroscope has an always-free plan with 50GB ingested per month and 14-day retention, plus a $19 monthly Pro platform fee; choose it when standalone continuous profiling with ingestion-based limits is preferable.

Verdict

Metoro is a strong choice for Kubernetes teams that need continuous CPU and memory profiling tied directly to operational context, especially across unmodifiable vendor workloads. Choose it for no-code eBPF coverage, Kubernetes filtering, and flexible deployment; look elsewhere when you need a broader monitoring suite, open-source-first deployment, or more predictable costs than per-node and excess-data billing provide.