Tech ONTAP Blogs

cancel
Turn on suggestions
Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.
Showing results for 
Search instead for 
Did you mean: 

Tech ONTAP Blogs

Harvest Speaks: New AI-Ready Observability Across All Your NetApp systems

AshishBajpai
NetApp
NetApp
‎2026-02-24 04:04 PM
2,185 Views

Harvest Speaks: New AI-Ready Observability Across All Your NetApp Systems


For years, NetApp® Harvest has been the quiet, open-source workhorse behind NetApp® observability, reliably collecting hardware metrics at scale. It exports these metrics as open-standard Prometheus metrics, delivering pre-built Grafana dashboards in an Apache 2.0 open-source stack. It was efficient, predictable, and intentionally unobtrusive.

 

But storage environments didn’t stay simple. Enterprises now work across NetApp ONTAP®, StorageGRID®, and E‑Series systems, often alongside infrastructure telemetry. As environments grew more heterogeneous, teams needed more than raw metrics—they needed a way to reason about their data across platforms, on the fly.

 

For the first time, Harvest communicates through a Model Context Protocol (MCP). It transforms telemetry into an AI-accessible layer that delivers intelligent, conversational, insights into your entire NetApp data estate, with a consistent experience across all three. The result is an AI-driven, conversational interface that queries, analyzes and interprets real time.

 

This new release marks a meaningful evolution of Harvest: from a metrics collector into the first unified, multi-platform observability foundation across the entire NetApp storage portfolio.

 

Harvest Speaks image 1 - Conversation.png

 

 If your storage estate were a city, Harvest would be the command center —maintaining real‑time awareness across ONTAP, StorageGRID, E‑Series, and Cisco environments, while an AI co‑pilot helps you navigate what matters most.

 

Unified Observability Across NetApp Storage


For customers running in mixed NetApp environments, as most do, Harvest now provides a single, vendor supported way to monitor ONTAP, StorageGRID, E-Series systems (preview), and Cisco environments, using a consistent metrics model. Performance, capacity, health, and hardware telemetry are collected using the same open standards and exported to your existing timeseries database—Prometheus, VictoriaMetrics, or InfluxDB.

 

And, yes, E-Series is now a first-class Harvest citizen, delivering the same observability experience that ONTAP and StorageGRID users have had for years.

 

  • Purpose-built Grafana dashboards for E-Series systems
  • Deep visibility into performance, capacity, health, and volume statistics
  • Open metrics output aligned to Prometheus standards — plug into your existing monitoring stack


The result is not a new monitoring silo, but a cleaner, more complete dataset that works with the tools teams already use. The purpose built Grafana dashboards remain available, but they are no longer the only way to extract insight.

 

Unified observability.png

 

Think of this phase as moving from ‘two instruments’ to a full orchestration.

From Dashboards to Direct Answers with Harvest MCP


The most significant addition included in this release is Harvest MCP, an implementation of the Model Context Protocol that sits directly on top of your existing metrics pipeline. Observability meets conversation.

 

Harvest MCP exposes your timeseries data through a structured, standards-based API that AI tools can query safely and consistently. Instead of building dashboards or writing PromQL, teams can ask natural-language, direct questions about your storage estate such as:

 

  • "How's my infrastructure doing right now?"
  • "Which systems are showing high latency in {xxx}?"
  • "Show me the top 5 ONTAP volumes by utilization in {xxx}."
  • "Analyze 30-day growth for my ONTAP aggregates."

MCP compatible assistants—such as Copilot or Claude Desktop—translate these questions into structured queries against your metrics store and return results in plain language.

 

Importantly, this does not replace dashboards or alerts. It augments them by making exploratory analysis faster and more accessible.

 

Dashboards to conversations.png

 

Architecture: Simple by Design

Harvest MCP does not introduce a new data pipeline. It builds directly on what you already run:

 

  • Harvest pollers collect metrics from ONTAP, StorageGRID, E‑Series, and supported Cisco environments.
  • Prometheus or VictoriaMetrics store the time‑series data.
  • Harvest MCP provides an AI‑accessible API layer over that dataset.
  • AI clients query the MCP endpoint using natural language.

 

If Harvest is already deployed, enabling MCP is a lightweight, additive step.

 

Getting Started with Harvest MCP

 

If your environment already exports Harvest metrics to Prometheus or VictoriaMetrics, deploying MCP requires running a single container. Once connected, any MCP‑compatible client can begin querying your storage telemetry immediately.

 

Step-by-step

 

  1. Run the Harvest MCP server
    docker run -d \
  --name harvest-mcp-server \
  -p 8082:8082 \
  --env HARVEST_TSDB_URL=http://your-prometheus:9090 \
  ghcr.io/netapp/harvest-mcp:latest \
  start --http --port 8082 --host 0.0.0.0​

    What this does

    1. Spins up the MCP API server

    2. Connects it directly to your TSDB

    3. Exposes a secure, standards-based Model Context Protocol interface

    4. Makes your telemetry discoverable and queryable by AI assistants

  2. Connect the MCP to your AI assistant
    a. Point any MCP-compatible client at your MCP server. See the integration guide for client-specific setup

  3. Ask a question- That’s it! Your storage telemetry is now conversational!

Once the MCP server is running, plug it into any MCP-compatible client- Copilot, Claude Desktop, or similar tools.

 

  • Example questions include:
    • “How healthy are my StorageGRID systems right now?”
    • “Which systems show performance issues in {xxx}?”
    • “Are there any Cisco environment health issues right now?”
    • “Which ONTAP volumes are approaching capacity limits?”
    • “What are the top five volumes by latency in {xxx} for the last 30 minutes?”
    • For additional examples, click here

 No new dashboards. No custom queries. Just direct access to your data with instant intelligent insights.

 

Closing thoughts


Our vision: Observability that thinks with you


Harvest MCP represents a shift in how teams interact with observability data. Metrics are still the foundation—but insight no longer depends on navigating dashboards or learning query languages.

 

NetApp’s broader MCP initiative (multiple MCP servers for different services) underscores that this is a strategy, not a sidecar. This new Harvest MCP is a part of a growing catalogue of MCP servers from NetApp. When storage data can answer questions directly, analysis becomes part of daily operations rather than a separate task.

 

Storage monitoring is no longer something you look at - it’s something you talk to.

And it talks back.


What’s next: Get involved and stay connected

  • Get involved and star Harvest GitHub repository

    • The most impactful ways to support the project is simply to star it

    • Starring helps signal its importance, increases visibility, and strengthens the ecosystem around the project

  • Join joining the communities in Discord and GetHub Discussions

    • It’s a terrific way to collaborate, learn, and stay informed

  • New to Harvest? Try Harvest MCP: 

    docker pull ghcr.io/netapp/harvest-mcp

     

 

All Community Forums
Public