> For the complete documentation index, see [llms.txt](https://developer.harness.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://developer.harness.io/resilience-testing/troubleshooting-and-resources/hce-vs-litmus.md). # LitmusChaos vs Harness Resilience Testing Compare open-source LitmusChaos with Harness Resilience Testing across deployment, fault coverage, testing disciplines, risk detection, and governance **LitmusChaos** is the CNCF-incubating, open-source chaos engineering project originally created by ChaosNative, which Harness later acquired. **Harness Resilience Testing** is built on that same chaos engine and is maintained by much of the same engineering team. This is not a comparison of two unrelated products. It is a comparison between the open-source engine on its own and the commercial platform built around it: the additional platform coverage, governance, testing disciplines, predictive risk detection, and AI-driven automation that Harness has layered on top since the project entered CNCF. The sections below compare the two on the dimensions a technical evaluation typically covers. A final section compares every named fault in the Harness Resilience Testing catalogue against the LitmusChaos published catalogue, fault by fault. Facts are current as of August 2026. Roadmap items are marked with their target general availability quarter; nothing on the roadmap is described as shipped. {% hint style="info" %} **Shared lineage** Harness Resilience Testing and LitmusChaos share a lineage. An existing Litmus deployment is an asset rather than a sunk cost when you evaluate Harness Resilience Testing. Harness Resilience Testing imports LitmusChaos experiment definitions unchanged, so an existing experiment library carries over on adoption. {% endhint %} *** ## What you will learn from this topic * **Origin and relationship:** How LitmusChaos and Harness Resilience Testing relate, and what carries over from an existing Litmus deployment. * **Platform comparison:** Deployment, fault coverage, testing disciplines, risk detection, AI, governance, observability, and commercial model. * **Fault catalogue:** A fault-by-fault comparison of Harness Resilience Testing's 215 named faults against LitmusChaos's published experiments. * **Evaluation guidance:** When each option is the better fit for your estate. *** ## Origin and relationship | Dimension | LitmusChaos (open source) | Harness Resilience Testing | | ------------------------ | ------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------- | | Creator / maintainer | Originally created by ChaosNative; now maintained as a CNCF Incubating project | Harness (acquired ChaosNative; the Resilience Testing chaos engine shares its lineage with LitmusChaos) | | Governance model | Community-governed, open contribution | Vendor-owned product roadmap with enterprise service level agreements | | Experiment compatibility | Native format | Imports LitmusChaos experiment definitions unchanged | | Hosted SaaS availability | LitmusChaos Cloud migrated onto Harness enterprise SaaS during 2026; a free tier is available there with capped experiment runs | Native SaaS and Self-Managed Platform (SMP, air-gapped) offerings, including the same free tier | *** ## Deployment and architecture | Dimension | LitmusChaos (open source) | Harness Resilience Testing | | --------------------- | -------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------- | | Deployment model | Self-hosted only, deployed via Helm onto a Kubernetes cluster | SaaS and Self-Managed Platform (SMP), including fully air-gapped install | | Multi-tenancy | No native concept of orgs or projects; relies on Kubernetes namespace and cluster boundaries | Native multi-org / multi-project structure | | Operational ownership | Customer operates and upgrades the control plane, database, and workflow engine | Harness operates the SaaS control plane; SMP ships as a versioned, supported appliance | | Release cadence | Monthly community releases | Vendor release train with change management and enterprise support | *** ## Fault and platform coverage | Dimension | LitmusChaos (open source) | Harness Resilience Testing | | -------------------------------- | ------------------------------------------------------------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------- | | Fault catalogue breadth | ChaosHub is community-curated and concentrated on Kubernetes pod and node disruption | 215 named faults across 8 platform families: Kubernetes 61, AWS 66, Linux 26, VMware/vSphere 25, Cloud Foundry/Tanzu 13, Azure 9, Windows Server 9, GCP 6 | | Kubernetes depth | Strong: pod, node, and container-level faults are the project's core focus | Inherits the same Kubernetes fault set, plus JVM and application-layer faults (SQL, MongoDB, Kafka, Solace) | | Non-Kubernetes hosts and clouds | Limited and community-contributed; not a first-class supported surface | Native agents for Linux and Windows hosts, with structured coverage across AWS, Azure, GCP, VMware/vSphere, and Cloud Foundry/Tanzu | | Custom / bespoke faults | Author custom experiments in the open-source, Go-based operator framework | Author custom faults in a private catalogue with no vendor engineering ticket; re-use an existing AWS FIS experiment as a native step | | Steady-state validation (probes) | HTTP, command, Kubernetes, and Prometheus probes | Same probe framework, extended with Datadog, Dynatrace, and New Relic probes | Go to [Chaos faults](/resilience-testing/chaos-testing/chaos-faults.md) to browse the full Harness Resilience Testing fault catalogue. *** ## Testing disciplines | Dimension | LitmusChaos (open source) | Harness Resilience Testing | | ----------------------------------- | ---------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------- | | Chaos testing | Yes: the project's sole discipline | Yes, on the same engine | | Load / performance testing | Not provided | Yes: traffic ramp profiles and SLO breach detection, in the same platform and pipeline as chaos | | Disaster recovery testing | Not provided | Yes: automated recovery time objective (RTO) and recovery point objective (RPO) failover validation | | Passive / predictive risk detection | Not provided | Yes: zero-config scan of deployment manifests, pipeline history, and infrastructure metadata; surfaces risk before any fault is injected | {% hint style="info" %} **Feature availability** Load testing, disaster recovery testing, and predictive risk detection may require feature flags on your account. Go to [Get started with load testing](/resilience-testing/load-testing/get-started.md), [Get started with DR testing](/resilience-testing/disaster-recovery-testing/get-started.md), and [Risks](/resilience-testing/shared-capabilities/risks.md) to confirm availability and contact your account administrator or Harness sales representative to enable them. {% endhint %} *** ## Predictive risk detection LitmusChaos has no static or predictive risk-detection capability at any level. Every row in this section is **Not provided** on the LitmusChaos side. For a buyer coming from LitmusChaos, this is the largest capability gap to close, not the fault catalogue. | Dimension | LitmusChaos (open source) | Harness Resilience Testing | | ----------------------------- | ------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Zero-config risk discovery | Not provided | Continuously scans deployment manifests, pipeline history, and infrastructure metadata for resilience gaps; no agent, no instrumentation, no code change | | Risk taxonomy | Not provided | 13 named dimensions across three severities: **FAIL** (single replica, no pod disruption budget, missing health probes, missing resource limits, single infra target); **WARN** (no horizontal pod autoscaler, no anti-affinity, no circuit breaker, missing resource requests, privileged container); **INFO** (no chaos coverage, no load coverage, stale deployment) | | Resilience Risk Score (RRS) | Not provided | FAIL × 10 + WARN × 4 + INFO × 1, with service-tier multipliers (Tier-0 × 2.0, Tier-1 × 1.5, Tier-2 × 1.0); per-service score bounded 0–200, pipeline-level score normalised 0–1000 | | Risk categories | Not provided | Availability, performance, dependency, configuration, and coverage in product messaging; the API adds recovery and security for 7 total | | Passive vs. active risk model | Not provided | Passive risks (probabilistic, from metadata; lifecycle DETECTED → EXPERIMENT SUGGESTED → RUNNING → CONFIRMED/CLEARED) versus active risks (raised when a probe actually fails, weighted roughly 2× a passive risk, and cannot be suppressed without a passing experiment) | | Risk dashboard and hub | Not provided | Single filterable view of every detected risk with severity badges, raw evidence, and a plain-language explanation of what was found | | CI-integrated risk detection | Not provided | Flags resilience regressions (missing resource limits, new untested dependencies, coverage drops) directly on pull requests, before they reach production | | Deployment gate on risk | Not provided | The PRD-Report API gates environment promotion on the resilience risk score, callable from any continuous delivery pipeline | | Automated experiment policies | Not provided | Automatically triggers a chaos or load experiment whenever a FAIL-severity risk is detected pre-production | | Risk suppression and audit | Not provided | Accept a known risk with a mandatory justification and an expiry (maximum 90 days); every detection, suppression, and status change is logged to an immutable audit trail | | Scan scale and performance | Not applicable | Full account scan of up to 500 services in 30 minutes, incremental scan in 5 minutes, false-positive rate under 10%, scales to 5,000 services / 50 projects / 20 integrations | | Data scope and retention | Not applicable | Reads deployment, pipeline, and infrastructure metadata only; never source code or customer data; scan metadata is not retained beyond a 72-hour window | Go to [Risks](/resilience-testing/shared-capabilities/risks.md) to understand how Harness Resilience Testing detects, scores, and confirms risks. *** ## AI and automation Roadmap items are marked with their target general availability quarter. Everything else in this section is shipped as of August 2026. | Dimension | LitmusChaos (open source) | Harness Resilience Testing | | ------------------------------- | ---------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------ | | Natural-language interface | LitmusChaos MCP Server, introduced in 2026, lets engineers query and learn about chaos experiments in natural language | AI agents embedded across the workflow; an MCP server for experiment create-read-update-delete is on the roadmap for Q4 2026 | | AI-generated experiments | Not provided | Test-generation agent converts a detected risk into a parameterised chaos experiment (shipped) | | AI recommendation of fixes | Not provided | Recommendation agent proposes a remediation alongside every detected risk (shipped) | | Post-run analysis | Manual review of probe results and dashboards | Post-run analysis agent (shipped) | | AI agents as a test target | Not provided | Chaos for AI Agents: model endpoint failure, tool-call timeout, degraded retrieval, rate limiting, silent model swap, cascading agent chains (roadmap, GA Q4 2026) | | Cascade / dependency-aware risk | Not provided | Graph-Aware Cascade Risk Detection (roadmap, Q4 2026) | | AI-generated DR tests | Not provided | Roadmap, Q1 2027 | Go to [AI Reliability Agent](/resilience-testing/shared-capabilities/ai/ai-reliability-agent.md) to configure AI-assisted experiment creation and remediation guidance. *** ## Governance, security, and compliance | Dimension | LitmusChaos (open source) | Harness Resilience Testing | | ------------------------- | --------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Authentication | No native single sign-on; inherits whatever authentication sits in front of the cluster | Security Assertion Markup Language (SAML), OAuth, and Lightweight Directory Access Protocol (LDAP) | | Authorisation | Kubernetes role-based access control only, enforced at the cluster level | Application-level custom role-based access control roles, scoped per project | | Two-factor authentication | Not provided | Yes | | Audit trail | Not provided as a product feature | Immutable audit trail covering risk detection, suppression, and status changes, positioned for Sarbanes-Oxley (SOX) and System and Organization Controls 2 (SOC 2) evidence | *** ## Observability and ecosystem | Dimension | LitmusChaos (open source) | Harness Resilience Testing | | -------------------------------------------------------- | ---------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------- | | Native metrics and dashboards | Prometheus metrics export (added 2026) with built-in dashboards | Native dashboards extended with cross-module views spanning chaos, load, disaster recovery, and risk results | | Application performance monitoring integration | Prometheus; further integration is left to the customer | Prometheus, Dynatrace, and other application performance monitoring probes wired directly into the automated pass/fail verdict, not just displayed | | Continuous integration / continuous delivery integration | Callable from any continuous integration pipeline via its API; no native continuous delivery product | Native step inside the Harness continuous delivery pipeline, or a REST API for other continuous delivery tooling | *** ## Community, support, and commercial model | Dimension | LitmusChaos (open source) | Harness Resilience Testing | | ----------------------- | ------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------- | | Cost of the software | Free and open source | Free entry tier with capped experiment runs; paid tiers priced per service, with unlimited experiment runs at higher tiers | | Support model | Community support via GitHub and Slack | Vendor service level agreement-backed enterprise support with a named customer success contact | | Enterprise-grade Litmus | Sold today as Harness's own Litmus enterprise distribution; there is no independent third-party enterprise offering | — | | Training | Community documentation and third-party courses | Harness University, plus the same third-party training | *** ## Chaos fault catalogue Harness Resilience Testing's published catalogue names 216 individual faults across eight platform families (the summary figure of **215** rounds the total in Harness materials). LitmusChaos's published catalogue totals approximately 47 named experiments. Of the rows below, 44 have a direct, identically-scoped LitmusChaos equivalent, marked **Yes** with the LitmusChaos experiment name. The remainder are application-layer, datastore-specific, or cloud-provider and operating-system extensions specific to Harness Resilience Testing, or currently have no publicly documented LitmusChaos equivalent. LitmusChaos's extensible operator framework means a determined team could author a custom equivalent for many rows marked **No** themselves. The tables below reflect what ships out of the box in each catalogue as of August 2026. ### Kubernetes (61 faults) **Pod lifecycle** | Fault | LitmusChaos (open source) | | -------------- | ------------------------- | | Pod delete | Yes: Pod Delete | | Pod scale | No | | Container kill | Yes: Container Kill | | Pod autoscaler | Yes: Pod Autoscaler | | Disk fill | Yes: Disk Fill | | Fs fill | No | **Pod resources** | Fault | LitmusChaos (open source) | | -------------- | ------------------------- | | Pod CPU hog | Yes: Pod CPU Hog | | Pod memory hog | Yes: Pod Memory Hog | | Pod IO stress | Yes: Pod IO Stress | **Pod network** | Fault | LitmusChaos (open source) | | ----------- | ---------------------------- | | Latency | Yes: Pod Network Latency | | Loss | Yes: Pod Network Loss | | Corruption | Yes: Pod Network Corruption | | Duplication | Yes: Pod Network Duplication | | Partition | Yes: Pod Network Partition | | Rate limit | Yes: Pod Network Rate Limit | **Pod DNS** | Fault | LitmusChaos (open source) | | --------- | ------------------------- | | DNS error | Yes: Pod DNS Error | | DNS spoof | Yes: Pod DNS Spoof | **Pod HTTP** | Fault | LitmusChaos (open source) | | ------------- | --------------------------- | | Latency | Yes: Pod HTTP Latency | | Status code | Yes: Pod HTTP Status Code | | Modify body | Yes: Pod HTTP Modify Body | | Modify header | Yes: Pod HTTP Modify Header | | Reset peer | Yes: Pod HTTP Reset Peer | **Pod API** | Fault | LitmusChaos (open source) | | ------------------------ | ------------------------- | | Block | No | | Latency | No | | Status code | No | | Modify body | No | | Modify header | No | | Modify response (custom) | No | **Pod IO** | Fault | LitmusChaos (open source) | | ------------------ | ------------------------- | | IO error | No | | IO mistake | No | | IO latency | No | | Attribute override | No | **Pod JVM** | Fault | LitmusChaos (open source) | | ---------------- | ------------------------------ | | CPU stress | Yes: Spring Boot CPU Stress | | Memory stress | Yes: Spring Boot Memory Stress | | Method latency | Yes: Spring Boot Latency | | Method exception | Yes: Spring Boot Exceptions | | Modify return | No | | Trigger GC | No | | SQL latency | No | | SQL exception | No | | Mongo latency | No | | Mongo exception | No | | Kafka latency | No | | Kafka exception | No | | Solace latency | No | | Solace exception | No | **App function** | Fault | LitmusChaos (open source) | | --------- | ------------------------- | | Error | No | | Latency | No | | Exception | No | **Redis** | Fault | LitmusChaos (open source) | | ----------------- | ------------------------- | | Cache expire | No | | Cache limit | No | | Cache penetration | No | **Time** | Fault | LitmusChaos (open source) | | ---------- | ------------------------- | | Time chaos | No | **Node** | Fault | LitmusChaos (open source) | | -------------------- | ------------------------- | | Drain | Yes: Node Drain | | Taint | Yes: Node Taint | | Restart | Yes: Node Restart | | CPU hog | Yes: Node CPU Hog | | Memory hog | Yes: Node Memory Hog | | IO stress | Yes: Node IO Stress | | Network latency | No | | Network loss | No | | Kubelet service kill | Yes: Kubelet Service Kill | ### Cloud Foundry / Tanzu (13 faults) **App** | Fault | LitmusChaos (open source) | | -------------- | ------------------------- | | App stop | No | | Container kill | No | | Route unmap | No | **Network** | Fault | LitmusChaos (open source) | | ----------- | ------------------------- | | Latency | No | | Loss | No | | Corruption | No | | Duplication | No | **JVM** | Fault | LitmusChaos (open source) | | ---------------- | ------------------------- | | CPU stress | No | | Memory stress | No | | Method latency | No | | Method exception | No | | Modify return | No | | Trigger GC | No | ### AWS (66 faults) **EC2** | Fault | LitmusChaos (open source) | | ------------------ | ------------------------- | | Stop by ID | Yes: EC2 Stop By ID | | Stop by tag | Yes: EC2 Stop By Tag | | CPU hog | No | | Memory hog | No | | IO stress | No | | Process kill | No | | Network latency | No | | Network loss | No | | DNS chaos | No | | HTTP latency | No | | HTTP status code | No | | HTTP modify body | No | | HTTP modify header | No | | HTTP reset peer | No | **EC2 Windows** | Fault | LitmusChaos (open source) | | --------------- | ------------------------- | | CPU hog | No | | Memory hog | No | | Network latency | No | | Network loss | No | | Process kill | No | | Blackhole | No | **EBS** | Fault | LitmusChaos (open source) | | ----------- | ------------------------- | | Loss by ID | Yes: EBS Loss By ID | | Loss by tag | Yes: EBS Loss By Tag | **Zone and network** | Fault | LitmusChaos (open source) | | ------------------------------------------------ | ------------------------- | | Availability zone blackhole | No | | Application Load Balancer availability zone down | No | | Classic Load Balancer availability zone down | No | | Network Load Balancer availability zone down | No | | Virtual private cloud route misconfiguration | No | **Data** | Fault | LitmusChaos (open source) | | ------------------------------------------- | ------------------------- | | Relational Database Service instance reboot | No | | Relational Database Service instance delete | No | | DynamoDB replication pause | No | **Elastic Container Service container** | Fault | LitmusChaos (open source) | | ---------------- | ------------------------- | | CPU hog | No | | Memory hog | No | | IO stress | No | | Network latency | No | | Network loss | No | | Volume detach | No | | HTTP latency | No | | HTTP status code | No | | HTTP modify body | No | | HTTP reset peer | No | | Agent stop | No | **Elastic Container Service task** | Fault | LitmusChaos (open source) | | ------------------------------- | ------------------------- | | Task stop | No | | Task scale | No | | Instance stop | No | | Network restrict | No | | Invalid container image | No | | Update container resource limit | No | | Update container timeout | No | | Update task role | No | | Fargate CPU hog | No | | Fargate memory hog | No | **Lambda** | Fault | LitmusChaos (open source) | | --------------------------- | ------------------------- | | Inject latency | No | | Inject status code | No | | Modify response body | No | | Update function memory | No | | Update function timeout | No | | Update role permission | No | | Delete function concurrency | No | | Function layer detach | No | | Block TCP connection | No | | Delete event source mapping | No | | Toggle event mapping state | No | **Identity and access management and Systems Manager** | Fault | LitmusChaos (open source) | | ----------------------------------------------------- | ------------------------- | | Resource access restrict | No | | Systems Manager chaos by ID | Yes: AWS SSM Chaos By ID | | Systems Manager chaos by tag | Yes: AWS SSM Chaos By Tag | | Generic Fault Injection Simulator experiment template | No | ### Azure (9 faults) **Compute and platform** | Fault | LitmusChaos (open source) | | ---------------------------------- | ------------------------- | | Azure Kubernetes Service node down | No | | Disk loss | Yes: Azure Disk Loss | | Instance stop | Yes: Azure Instance Stop | | Instance CPU hog | No | | Instance memory hog | No | | Instance IO stress | No | | Web app stop | No | | Web app access restrict | No | | Service bus queue state change | No | ### Linux hosts and bare metal (26 faults) **Stress** | Fault | LitmusChaos (open source) | | -------------- | ------------------------- | | CPU stress | No | | Memory stress | No | | Disk fill | No | | Disk IO stress | No | | Fs fill | No | **Network** | Fault | LitmusChaos (open source) | | ----------- | ------------------------- | | Latency | No | | Loss | No | | Corruption | No | | Duplication | No | | Rate limit | No | | DNS error | No | | DNS spoof | No | **Process and time** | Fault | LitmusChaos (open source) | | --------------- | ------------------------- | | Process kill | No | | Service restart | No | | Time chaos | No | **JVM** | Fault | LitmusChaos (open source) | | ---------------- | ------------------------- | | CPU stress | No | | Memory stress | No | | Method latency | No | | Method exception | No | | Modify return | No | | Trigger GC | No | **API** | Fault | LitmusChaos (open source) | | ------------- | ------------------------- | | Block | No | | Latency | No | | Status code | No | | Modify body | No | | Modify header | No | ### Windows Server (host agent) (9 faults) **Native Windows agent** | Fault | LitmusChaos (open source) | | ------------------- | ------------------------- | | CPU stress | No | | Memory stress | No | | Disk stress | No | | Process kill | No | | Network latency | No | | Network loss | No | | Network corruption | No | | Network duplication | No | | Blackhole chaos | No | ### VMware / vSphere (25 faults) **Linux guest** | Fault | LitmusChaos (open source) | | -------------------- | ------------------------- | | CPU hog | No | | Memory hog | No | | IO stress | No | | Network latency | No | | Network loss | No | | Network rate limit | No | | DNS chaos | No | | HTTP latency | No | | HTTP reset peer | No | | HTTP response modify | No | | Process kill | No | | Service stop | No | **Windows guest** | Fault | LitmusChaos (open source) | | ------------------- | ------------------------- | | CPU hog | No | | Memory hog | No | | Disk stress | No | | Process kill | No | | Service stop | No | | Time chaos | No | | Blackhole chaos | No | | Network latency | No | | Network loss | No | | Network corruption | No | | Network duplication | No | **vCenter** | Fault | LitmusChaos (open source) | | ------------------- | ------------------------- | | VM poweroff | Yes: VM Poweroff | | VM poweroff by name | No | ### Google Cloud (6 faults) **Compute and SQL** | Fault | LitmusChaos (open source) | | ------------------------- | ---------------------------------- | | VM instance stop | Yes: GCP VM Instance Stop | | VM instance stop by label | Yes: GCP VM Instance Stop By Label | | VM disk loss | Yes: GCP VM Disk Loss | | VM disk loss by label | Yes: GCP VM Disk Loss By Label | | VM service kill | No | | SQL instance failover | No | *** ## Summary For a team already running LitmusChaos in production, adopting Harness Resilience Testing is closer to a continuation than a migration. The underlying fault engine, probe model, and experiment definitions carry across unchanged. What changes is everything built around that engine: multi-cloud and host fault coverage (215 named faults against LitmusChaos's roughly 47), enterprise authentication and audit, load and disaster-recovery testing on the same pipeline as chaos, and a predictive risk-detection layer that has no LitmusChaos equivalent at any level. LitmusChaos on its own remains a reasonable choice for a single Kubernetes cluster, no compliance requirement, and an engineering team willing to operate and upgrade the control plane itself. Harness Resilience Testing is the better fit once the estate spans more than one platform or team, carries a governance and audit requirement, or needs to know where the risk is before spending engineering time proving it experimentally, and it does so without discarding any LitmusChaos investment already made. *** ## Next steps * [Overview](/resilience-testing/new-to-resilience-testing/overview.md): Understand the three pillars of Harness Resilience Testing and how they work together. * [Chaos faults](/resilience-testing/chaos-testing/chaos-faults.md): Browse the full Harness Resilience Testing fault catalogue. * [Risks](/resilience-testing/shared-capabilities/risks.md): Learn how Harness Resilience Testing detects and scores resilience risks before you run an experiment. * [What's supported](/resilience-testing/troubleshooting-and-resources/whats-supported.md): Confirm platform, infrastructure, and feature support for your environment. {% @harness-feedback/feedback %}