Buyer’s Guide - August 2026

The 7 best Ceph monitoring tools, ranked

Q: How often should Ceph metrics be collected?

Per-second collection is ideal for catching short-lived OSD latency and IOPS spikes that longer intervals miss. Netdata collects per second by default, Prometheus scrapes at a 15-second default, and Zabbix and Checkmk poll at tens of seconds to minutes per item. In practice, a 15-second scrape averages away sub-minute spikes, and minute-level polling never sees them at all.

Q: How do I set up alerting for Ceph?

On the standard stack you write Prometheus rules and route them through Alertmanager; the Ceph Dashboard can manage silences and list alerts on top of that. Netdata ships a Ceph capacity-utilization alert and adds ML-based anomaly detection on Ceph metrics without rule files. Zabbix ships built-in triggers and Datadog ships service checks covering OSD down, OSD full, and PG degradation.

Q: Can I monitor Ceph in Kubernetes with Rook?

Yes. Rook exposes Ceph metrics via the Prometheus manager module and a ServiceMonitor, so you can scrape them with the cluster Prometheus and import the Grafana dashboards. Netdata auto-discovers Ceph Manager instances inside Docker containers, which covers Rook deployments with no extra configuration beyond credentials. Zabbix and Checkmk can monitor Rook Ceph hosts through their agents.

Ceph already ships the hard part: the manager modules expose rich cluster, OSD, pool, and placement group metrics. What separates the tools on this list is what they do with those metrics: collection resolution, dashboards, alerting, and whether you get one pane across many clusters. We ranked seven options on Ceph metric depth, time to value, alerting, fleet visibility, and cost shape.

Start with Netdata (free) Jump to rankings

Why this list exists

Ceph monitoring is an unusual category because the data problem is already solved. The mgr prometheus module exposes cluster metrics, service discovery, and RBD image stats on port 9283; ceph-exporter and the admin socket cover per-daemon performance counters; the Ceph Dashboard embeds Grafana panels out of the box. Every serious tool on this list reads from the same sources. So the buying decision is not “who can see my Ceph cluster” but “who turns those metrics into answers with the least assembly and the best resolution.”

The two mistakes we see most often: assuming the built-in Ceph Dashboard is enough for production (it is single-cluster, and its monitoring depth depends on the Prometheus and Grafana stack you deploy alongside it), and assuming Prometheus plus Grafana is the only serious option (it is the reference architecture, but you assemble and operate exporters, scrape configs, retention, HA, and dashboards yourself). Agent-based tools close that gap with per-second collection and zero stack assembly.

Three dimensions decide the outcome for most Ceph fleets:

Collection resolution. The Prometheus default scrape interval is 15 seconds. Short-lived OSD latency spikes and IOPS bursts average out at that cadence. Per-second collectors catch them; minute-level pollers never see them.
Assembly cost. The full open source stack (mgr module, ceph_exporter, scrape configs, Grafana dashboards, Alertmanager rules, long-term storage) is a project. Prebuilt collectors, dashboards, and alerts compress that to minutes.
Fleet shape. A single Rook cluster and a dozen Proxmox Ceph clusters are different problems. Some tools are single-cluster by design; others give one pane across every node.

A note on pricing: we do not quote competitor list prices. Observability pricing changes often, and the number that matters is how the bill grows with your fleet: per host, per service, or per GB ingested. We describe each vendor’s pricing shape and link the official pricing page so you can check current figures. For operator-level setup guidance, our Ceph monitoring runbooks walk through the metrics that matter and how to collect them.

We assembled the shortlist from the tools Ceph operators actually run: the reference open source stack Ceph itself documents, the agent-based platforms with maintained Ceph integrations, the one SaaS with a real built-in Ceph check, and the built-in dashboard every cluster already has. We excluded generic infrastructure tools with no maintained Ceph integration.

The two heaviest criteria are Ceph metric depth and time to value, because they capture the real tradeoff in this category: the deepest raw metric access (the full per-daemon counter dump) lives in the stack with the highest assembly cost, while the fastest setups expose a curated metric set. Alerting, fleet-wide visibility, and cost shape round out the scoring; ecosystem and support act as tiebreakers.

Tester credit

Compiled by the Netdata team - Updated August 12, 2026

Sources verified

Scoring criteria

Ceph metric depth and granularity 25%
Coverage of cluster health, OSDs, pools, PGs, capacity, and per-daemon counters, and the collection resolution.
Time to value and ease of setup 20%
Auto-discovery and prebuilt dashboards versus assembling exporters, scrape configs, and dashboards.
Alerting and anomaly detection 15%
Built-in Ceph alerts and ML anomaly detection versus hand-written rules.
Fleet-wide and multi-cluster visibility 15%
One pane for many clusters, or single-cluster by design.
Cost and pricing model 10%
Open source self-hosted, per-node, per-host, or usage-based; what grows the bill.
Ecosystem and integrations 10%
Compatibility with the mgr prometheus module, ceph_exporter, Grafana, and Rook.
Support and maintenance 5%
Vendor support, community health, and how actively the Ceph integration is maintained.

Vendor 01 / 07 · #netdata

Netdata

Real-time, per-second Ceph cluster monitoring with auto-discovery, built-in dashboards, and ML-based anomaly detection.

Best for

Ceph admins who want per-second visibility without building a Prometheus+Grafana stack
Fleets mixing Ceph with Proxmox, Kubernetes, or general Linux infrastructure
Teams that want ML-assisted anomaly detection and root-cause analysis out of the box

Pricing

Per-node pricing; Netdata Cloud Business starts at $4.5/node/month on annual plans
Per-node price decreases as node count grows, which suits large Ceph fleets
Agents are open source (AGPLv3) and free to run self-hosted
Free Cloud tier exists for small fleets

Pros

Collects Ceph cluster, OSD, and pool metrics every second by default via the Ceph Manager REST API, versus the 15-second Prometheus scrape default
Auto-discovers Ceph Manager instances on localhost (port 8443) and inside Docker containers, which covers Rook deployments
Covers cluster status, OSD up/down/in/out, PG states, capacity utilization, and per-OSD and per-pool IOPS, throughput, and latency
Ships a Ceph capacity-utilization alert and ML-based anomaly detection on Ceph metrics with no rule writing
Netdata Cloud adds fleet-wide views, custom dashboards, and AI-assisted troubleshooting across all your Ceph nodes
Open source agent with 900+ integrations, so Ceph sits alongside hosts, VMs, and Kubernetes in one tool

Where teams pair it

The Ceph collector is Linux-only
It exposes a curated 27-metric set rather than the full per-daemon performance counter dump; teams that need raw counter depth pair it with the Prometheus mgr module

Verdict

Netdata is the fastest path from a running Ceph cluster to actionable visibility. The per-second default collection interval catches OSD latency and IOPS spikes that a 15-second scrape averages away, and auto-discovery plus prebuilt dashboards and alerts mean there is no exporter, scrape config, or dashboard assembly to schedule. ML anomaly detection on top of Ceph metrics surfaces the OSD or pool behaving oddly before it trips a threshold. The honest caveat: if your workflow depends on the full per-daemon counter dump from the mgr prometheus module, you will still want Prometheus alongside it, and deeper Ceph alert rules beyond the shipped capacity alert are DIY. For most Proxmox, Rook, and bare-metal fleets, it is the right first tool.

Start free -> Netdata Ceph monitoring 101 ->

Vendor 02 / 07 · #prometheus

Prometheus + ceph_exporter

The de facto standard open source metrics stack for Ceph, scraping the built-in mgr prometheus module and ceph_exporter.

Best for

Teams already standardized on Prometheus for the rest of their infrastructure
Ceph admins who want the full per-daemon performance counter set
Kubernetes and Rook deployments where Prometheus is the default

Pricing

Open source (Apache-2.0) and self-hosted: no license fee, but you run and operate it
Cost grows with metric cardinality, retention, and the hardware and ops time to run the stack
Long-term storage and HA typically add components such as Thanos or Mimir

Pros

Ceph ships a native mgr prometheus module (default port 9283) exposing cluster metrics, service discovery, and RBD image stats
digitalocean/ceph_exporter (default port 9128) adds per-daemon performance counters via the Ceph admin socket
The deepest raw Ceph metric access of any option on this list
Huge ecosystem: Alertmanager, community Grafana dashboards (7050, 2842), and Rook exposes a ServiceMonitor
The reference architecture Ceph’s own documentation describes

Cons

You assemble and operate everything yourself: exporters, scrape configs, retention, HA, dashboards, and alert rules
The default 15-second scrape interval misses sub-minute latency and IOPS spikes
No built-in dashboards or alerting UI; requires Grafana and Alertmanager setup
High cardinality and long retention need extra components and ongoing capacity planning

Verdict

If you want every counter Ceph can emit, this is where you end up, because the mgr module and ceph_exporter together expose what no curated integration matches. It is also the natural default in Rook environments where Prometheus already runs. The cost is operational, not licensing: someone on your team owns scrape configs, dashboard imports, rule files, and the storage layer, forever. Teams that standardize on it get a superb foundation; teams that just want Ceph answers should weigh the assembly hours against agent-based options.

Read full Netdata vs Prometheus -> Prometheus site ->

Vendor 03 / 07 · #grafana

Grafana

The standard visualization layer for Ceph metrics, with community dashboards and a managed cloud option.

Best for

Teams that want polished, shareable Ceph dashboards on top of Prometheus
Organizations already using Grafana for other observability
Buyers who want a managed option (Grafana Cloud) without running the stack

Pricing

Grafana OSS is open source (AGPLv3) and self-hosted; you operate it
Grafana Cloud has a free tier and usage-based Pro pricing
The cloud bill grows with active series, log and trace volume, and retention

Pros

Hundreds of community Ceph dashboards (Ceph Clusters Overview 7050, Ceph Cluster 2842) importable in one click
Ceph’s own documentation ships Grafana dashboards, with Prometheus as the data source
Rich alerting, annotations, and multi-cluster comparison views
The visualization layer the Ceph Dashboard itself embeds

Cons

Not a collector: it needs Prometheus (or another data source) and the Ceph exporters behind it
Community dashboards are community-maintained and can lag Ceph releases
Cloud usage-based pricing can grow with series count and retention

Verdict

Grafana is the best dashboard layer for Ceph, full stop, and the fact that the Ceph Dashboard embeds it tells you where the ecosystem landed. But it is half a stack: nothing reaches those panels until you stand up Prometheus, the exporters, and their scrape configs. Treat it as the presentation tier of the Prometheus entry above rather than a standalone Ceph monitoring choice, or as the managed front end via Grafana Cloud if you want someone else to run the backend.

Read full Netdata vs Grafana -> Grafana pricing ->

Vendor 04 / 07 · #zabbix

Zabbix

Open source enterprise monitoring platform with an official agent-based Ceph template.

Best for

Enterprises that want a full self-hosted monitoring platform with built-in alerting
Teams already running Zabbix for servers and network devices who want Ceph in the same tool
Buyers who prefer agent-based collection over scraping

Pricing

Open source (GPLv2+) with no license fee; you run and operate the server, database, and agents
Paid support subscriptions (Silver through Global) for enterprises that want a vendor behind it
A managed Zabbix Cloud option exists if you do not want to self-host the server

Pros

Official ‘Ceph by Zabbix agent 2’ template with no external scripts, collecting via the Ceph mgr restful module API
Covers cluster status, OSD up/down/fill, PG states, pool usage, bandwidth, IOPS, and OSD apply/commit latency
Built-in triggers for cluster ERROR/WARNING, OSD down, and OSD full/near-full
Template versions track Zabbix 5.0 through 7.4; tested against Ceph 14.2

Cons

Requires the Ceph mgr restful module with extended mon capabilities for the agent user
Polling-based collection at per-item intervals, coarser than per-second agents
The template covers cluster, OSD, and pool views but not the full per-daemon counter set

Verdict

Zabbix is the strongest complete-platform alternative on this list: one self-hosted system for Ceph, servers, and network gear, with a maintained official template and real Ceph triggers out of the box. For Zabbix shops it is an easy yes. For everyone else, the tradeoff is polling resolution (tens of seconds to minutes per item, so short spikes slip through) and the operational weight of running the Zabbix server and database to monitor a storage cluster that already has enough moving parts.

Read full Netdata vs Zabbix -> Zabbix subscriptions ->

Vendor 05 / 07 · #checkmk

Checkmk

Open-core monitoring platform that ships a maintained Ceph agent plugin for status, pools, and OSD checks.

Best for

Mixed infrastructure teams that want Ceph plus servers, VMs, and network in one platform
Organizations that prefer agent-based checks with a GUI-driven rule system
Buyers wanting a managed SaaS option (Checkmk Cloud)

Pricing

Open-core: a free Community edition capped by service count, and paid Pro, Ultimate, and Cloud editions
Editions are sized by monitored services, custom metrics, and synthetic tests
The bill grows with service count and custom metrics rather than host count

Pros

Ships the widely used Ceph statistics plugin (Robert Sander’s mk_ceph), integrated since Checkmk 2.4
Checks cover Ceph status, pools, classes, and OSD BlueFS devices (slow device, database, write-ahead log)
Agent plugin deployable via the Checkmk bakery ruleset
Mature alerting, dashboards, and reporting built on top of the checks

Cons

The Ceph plugin requires python and the third-party rados library on the monitored host
Check-based collection runs at minute-level granularity, not per-second
The Community edition is capped by service count; deeper Ceph metrics need paid editions

Verdict

Checkmk’s Ceph plugin is genuinely maintained and covers the operational basics well, including the BlueFS device checks most tools skip. It makes the most sense when Ceph is one line item in a broader Checkmk deployment, where the bakery rollout and rule engine already exist. As a Ceph-only choice it is harder to justify: minute-level checks will not show you a latency spike, and the service-count pricing model punishes fleets that monitor many pools and OSDs in detail.

Read full Netdata vs Checkmk -> Checkmk pricing ->

Vendor 06 / 07 · #datadog

Datadog

SaaS observability platform with a built-in Ceph integration for pools, OSDs, and I/O performance.

Best for

Organizations already standardized on Datadog for infrastructure and APM
Teams that want Ceph metrics alongside Kubernetes, cloud, and application monitoring in one SaaS
Buyers who want zero self-hosted monitoring infrastructure

Pricing

SaaS per-host subscription tiers plus usage-based add-ons
The bill grows with host count, custom metrics, and ingested log and trace volume
A limited free tier is capped by host count and retention

Pros

The Ceph check ships inside the Datadog Agent with no extra install, reporting pool, OSD, and cluster metrics
Service checks for overall status, OSD down/full, PG degraded/damaged, and slow or stuck requests
Log collection from /var/log/ceph/*.log alongside metrics
Red Hat documents Datadog as a supported Ceph monitoring option

Cons

Per-host SaaS pricing makes large Ceph fleets expensive; every storage node adds to the bill
The agent check runs the ceph CLI on each host and needs sudo or ceph access configuration
The integration is a curated metric set, not the full per-daemon counter dump

Verdict

Datadog’s Ceph integration is real and useful: the service checks for OSD down, PG degradation, and stuck requests are exactly the conditions you want paged. If Datadog is already your platform, enabling it is a configuration exercise, not a project. The problem is the pricing shape. Ceph clusters are node-dense by design, and per-host billing plus custom-metric add-ons turn a 30-node storage fleet into a line item finance will ask about. For dedicated storage fleets, a per-node tool with a decreasing unit price or a self-hosted stack usually wins on total cost.

Read full Netdata vs Datadog -> Datadog pricing ->

Vendor 07 / 07 · #ceph-dashboard

Ceph Dashboard

The web-based management and monitoring UI built into Ceph itself, with embedded Grafana dashboards.

Best for

Ceph admins who want zero extra tooling for day-to-day health checks
Small clusters where a second monitoring platform is overkill
Teams that want management (OSDs, pools, RBD, CephFS) and monitoring in one UI

Pricing

Part of Ceph (open source, LGPL) and self-hosted; no separate license or subscription
The real cost is the operational overhead of the Prometheus and Grafana stack it depends on for monitoring depth

Pros

Ships with Ceph as a manager module and is enabled with one command
Monitors cluster health, hosts, monitors, OSDs, pools, PGs, capacity, and performance counters
Embeds Grafana dashboards backed by Prometheus for utilization, IOPS, latency, and throughput
Device management with SMART data and drive health predictions
Prometheus alerting integration: silences, alert listing, and Alertmanager

Cons

Monitoring depth depends on the Prometheus and Grafana stack you deploy alongside it
Single-cluster focus; no fleet-wide multi-cluster view or long-term analytics
Not a standalone product; you get it only as part of Ceph

Verdict

Every Ceph admin already has this, and for daily health checks it earns its place: the device management view with SMART data and drive failure prediction is something no external tool replicates from metrics alone. The mistake is treating it as your production monitoring strategy. It watches one cluster at a time, its graphs inherit the 15-second Prometheus scrape interval, and long retention and advanced alerting live elsewhere. Use it as the built-in baseline it is, then pick one of the tools above for alerting, history, and fleet views.

Ceph Dashboard docs -> Ceph project site ->

Frequently Asked Questions

What is the best way to monitor a Ceph cluster?

The best way depends on your fleet size and team: the built-in Ceph Dashboard handles day-to-day health checks on a single cluster, while agent-based tools like Netdata give faster time-to-value with per-second collection, and the Prometheus plus Grafana stack offers the deepest raw metrics at higher assembly cost. SaaS options like Datadog fit teams already on that platform. The core tradeoff is resolution and setup effort versus raw metric depth.

What Ceph metrics should I monitor?

Start with the cluster health state (HEALTH_OK, HEALTH_WARN, HEALTH_ERR) and OSD up/in status with per-OSD space. Track placement group states, watching for anything that is not active+clean: degraded, misplaced, or unfound PGs. Add capacity utilization and nearfull/full ratios, then OSD apply and commit latency, IOPS, and throughput for performance work. Every tool on this list covers these; they differ on resolution and how much assembly it takes.

Does Ceph have built-in monitoring?

Yes. The Ceph Dashboard manager module provides a web UI with embedded Grafana dashboards. The mgr prometheus module (default port 9283) exposes cluster metrics, service discovery, and RBD image stats, and the ceph-exporter daemon covers per-daemon performance counters. Built-in monitoring is single-cluster, and its monitoring depth depends on the Prometheus and Grafana stack you deploy alongside it.

How do I monitor Ceph with Prometheus?

Enable the mgr prometheus module and scrape port 9283 at the default 15-second interval. Add the digitalocean ceph_exporter (port 9128) for per-daemon performance counters via the admin socket. Import community Grafana dashboards such as 7050 (Ceph Clusters Overview) and 2842, and use Alertmanager for alerting. On Kubernetes, Rook exposes a ServiceMonitor so the cluster Prometheus picks up Ceph metrics automatically.

How often should Ceph metrics be collected?

How do I set up alerting for Ceph?

Can I monitor Ceph in Kubernetes with Rook?

The 7 best Ceph monitoring tools, ranked

How we evaluated Ceph monitoring tools

Netdata

Prometheus + ceph_exporter

Grafana

Zabbix

Checkmk

Datadog

Ceph Dashboard

Frequently Asked Questions

Book Your Free Demo