Every Grafana and Prometheus book from a major publisher went to press before the releases you’ll install today. Prometheus went to 3.0 in November 2024, and Grafana has shipped three major versions since most of these books were printed. So the best Grafana books and Prometheus books in 2026 are the ones whose ideas outlast the screenshots. You still learn PromQL and dashboard design faster from a good book than from release notes, once you know which examples need adjusting.
Seven books made this list: three Grafana titles from first dashboard to the full LGTM stack, three Prometheus books from first scrape to Thanos-scale fleets, and one for network engineers. For OpenTelemetry and observability theory, our observability books list covers that ground. We checked every edition and price against Amazon’s catalog and read each publisher’s code repository or contents page in October 2026, against Prometheus 3.15 and Grafana 13.2.
Choose a book by the job in front of you
Start from what you run, not from the most famous cover. Prometheus: Up & Running is the strongest Prometheus reference here, but it is a Prometheus book, so a reader who mostly builds dashboards should start with a Grafana title instead:
- Best Grafana book overall: Observability with Grafana
- Best Prometheus book overall: Prometheus: Up & Running, 2nd Edition
- Best first Grafana book: Learn Grafana 10.x, 2nd Edition
- Best for scaling Prometheus: Mastering Prometheus
- Best for network engineers: Modern Network Observability
The other two fill narrower gaps. Hands-On Monitoring and Alerting with Prometheus is the only publisher title here printed after Prometheus 3.0, and a short route to working alert rules. End-to-End Observability with Grafana is the cheapest book here, with chapters on the blackbox exporter and synthetic monitoring that the other Grafana picks don’t have. Most readers need two books from this list: one Prometheus book for the metrics and alerting side, and one Grafana book for the dashboards your team will actually look at.
Check what each book was written for
We started from the publishers’ catalogs rather than Amazon’s best-seller pages. A catalog search for “Prometheus” returns Greek tragedy, Frankenstein and a wave of 2026 self-published titles, many under 100 pages, long before it reaches a monitoring book. Our Amazon catalog search covered Packt, O’Reilly, Manning, Apress, BPB and AVA. Every pick on this page is a print edition sold and shipped by Amazon when we checked, and each cover matches the exact edition we link.
Then we looked at what each book was written against. Packt publishes its book code on GitHub, and those repositories pin real versions: a kube-prometheus-stack chart, a Grafana image tag, an OpenTelemetry Collector chart. Where a publisher gives no code, we used its table of contents. We did not work through every chapter of every book, so when we describe what a book covers, that comes from the publisher’s contents page, the code repository or the listing, and we say so.
A few well-known titles are still on sale but were left out on purpose, because each has a newer replacement on this page:
- Hands-On Infrastructure Monitoring with Prometheus (Packt, 2019)
- Learn Grafana 7.0 (Packt, 2020)
- Getting Started with Grafana (Apress, 2022)
- Prometheus: Up & Running, first edition (O’Reilly, 2018)
Observability for Kubernetes Applications from BPB treats tools as interchangeable. It fits the broader observability list better than this one, so it gets no section below.
Compare the 7 Grafana and Prometheus books
The “Lab versions” column is what the book’s own code or lab files pin, which is the version its screenshots and outputs will match. “Not stated” means neither the publisher page nor any code we found names one:
| Book | Publisher, year | Pages | Focus | Lab versions | Level |
|---|---|---|---|---|---|
| Observability with Grafana | Packt, 2024 | 356 | Loki, Grafana, Tempo, Mimir with OpenTelemetry | Kubernetes 1.26, OTel Collector chart 0.73.1, Grafana Cloud | Intermediate |
| Prometheus: Up & Running, 2nd Ed | O’Reilly, 2023 | 415 | Prometheus server, PromQL, exporters, Alertmanager | Not stated (Prometheus 2.x era) | Beginner to advanced |
| Learn Grafana 10.x, 2nd Ed | Packt, 2023 | 542 | Dashboards, panels, templating, auth | Grafana 10.x in the text; compose files pull latest | Beginner |
| Mastering Prometheus | Packt, 2024 | 310 | Scaling, Thanos, Mimir, rule testing, SLOs | kube-prometheus-stack 47.0.0 (Prometheus v2.44.0) | Advanced |
| Hands-On Monitoring and Alerting with Prometheus | AVA, 2025 | 212 | Exporters, relabeling, PromQL, Alertmanager | Not stated | Beginner to intermediate |
| Modern Network Observability | Packt, 2024 | 506 | Telegraf, Prometheus, Loki, Grafana for networks | Prometheus v2.52.0, Grafana 10.4.4, Loki 3.1.1 | Intermediate |
| End-to-End Observability with Grafana | BPB, 2023 | 332 | Dashboards, Loki, synthetic and Kubernetes monitoring | Not stated | Beginner to intermediate |
Every book in that table that pins a version pins Prometheus 2.x or Grafana 10, which is the honest state of the shelf in 2026. That does not make the books wrong, but it does mean you should run the labs on the version the book pins first, then upgrade once the concept has clicked. The last section of this guide lists what changes when you do.
1. Observability with Grafana
Rob Chapman and Peter Holmes wrote the Grafana book to buy if you run more than dashboards. It covers the whole Grafana Labs stack, Loki for logs, Tempo for traces, Mimir and Prometheus for metrics, and it instruments a demo application with OpenTelemetry so you query its logs, metrics and traces in LogQL, PromQL and TraceQL. Chapman lists Grafana among his former employers, which helps on the parts of the stack that the product documentation explains least.
Packt’s code repository shows how the labs run. You install the OpenTelemetry Collector with Helm, pinned to chart version 0.73.1, on a Kubernetes 1.26 cluster, and the collector ships traces, metrics and logs to Grafana Cloud using credentials you create under Access Policies in the Cloud portal. That makes the setup quick, because you don’t deploy Loki, Tempo and Mimir yourself, but it also means the hands-on chapters assume a Grafana Cloud account. Later chapters move on to alerting and incident management, CI/CD pipeline monitoring, cost control and how to design an observability platform for a whole organization.
The book came out in January 2024, four months before Grafana 11, so its screenshots predate Grafana 11 and the Drilldown apps that became generally available in Grafana 12. If you want to self-host the stack instead of using Grafana Cloud, pair it with our guides to install Grafana Loki and install Grafana Mimir, then point the book’s collector config at your own endpoints. It ran about $40 against a $51.99 list price when we checked. Check the current price on Amazon.
2. Prometheus: Up & Running, 2nd Edition
This is still the Prometheus book. Julien Pivotto is a current Prometheus maintainer, and Brian Brazil is among the ten most prolific committers in the Prometheus server’s history. The explanations of how scraping, the data model and PromQL behave come from people who wrote that code. If you buy one book about Prometheus itself, buy this one.
O’Reilly published the second edition in May 2023 at 415 pages. According to the book’s preface, as excerpted on its Amazon page, it adds the PromQL functions, service discovery providers and Alertmanager receivers that arrived after the 2018 first edition, plus a new chapter on server-side security such as TLS. That last chapter matters more than it sounds, because a Prometheus left open on port 9090 with no TLS or authentication is a common mistake on self-hosted setups. The book takes you from instrumenting code and running exporters through PromQL, recording rules and alerting, which is the order you need when you install Prometheus on Ubuntu or Debian for the first time.
Its limit is the calendar. Prometheus 3.0 shipped 18 months after this edition, so the book does not cover UTF-8 metric names, the left-open range selectors or the renamed holt_winters function, and it predates native histograms going stable. Read it with the migration table at the end of this guide open. Amazon listed limited stock at about $43 against a $65.99 list price when we checked, and the 2018 first edition is still for sale under the same title, so confirm the listing says second edition. Check the current price on Amazon.
3. Learn Grafana 10.x, 2nd Edition
Eric Salituro’s book is the gentlest start on this list and the one to hand someone who has never opened Grafana. At 542 pages it goes panel by panel: time series visualization, transformations, the query inspector, annotations, template variables, dashboard links and sharing, then user authentication through Okta, Google and GitHub, and cloud monitoring. It uses Prometheus, InfluxDB and other sources as data rather than as subjects, so you learn Grafana without first learning a query language in depth.
One detail in the code repository is easy to miss. The chapter Docker Compose files use grafana/grafana:${GRAF_TAG-latest} and prom/prometheus:${PROM_TAG-latest}, so unless you set those variables, the labs pull today’s Grafana 13 while the book’s screenshots show 10.x. The repository’s Makefiles call the older hyphenated docker-compose. Skip make up and run GRAF_TAG=10.4.4 docker compose up -d in the chapter folder if you want the screens to match. PROM_TAG also defaults to latest, so the same lab pulls Prometheus 3 unless you pin it too. Our guide to install Grafana on Ubuntu and Debian covers the current release when you move to a real server.
Experienced users will find it slow. If you already build dashboards with variables and alert rules, most of its 542 pages cover ground you know, so one of the books above will teach you more per hour. Amazon also still sells the 2020 Learn Grafana 7.0 under a similar title, so check the cover says 10.x. About $44 when we checked. Check the current price on Amazon.
4. Mastering Prometheus
Per his author bio, William Hegedus manages an SRE team at Akamai that runs a global fleet of Prometheus servers. Mastering Prometheus is about what happens when one server stops being enough. The repository says a Prometheus background is assumed, so the book spends little time on installation and a lot on the TSDB data model, cardinality, custom service discovery and remote storage with Mimir, VictoriaMetrics or Thanos.
The code repository shows the operational habits the book teaches, chapter by chapter:
- chapter 10: a full Thanos deployment generated from jsonnet, with storage in Linode Object Storage set up through
linode-cli - chapter 12: alert and recording rules tested in CI with
pint,promtoolandamtool - chapter 13: SLO alerts generated with Pyrra
Those rule-testing workflows are worth copying into your own repository even if you never touch Thanos, because a broken alert rule is silent until the night you need it. Our guide to set up Prometheus Alertmanager is a good base to apply them to.
The labs install kube-prometheus-stack chart 47.0.0, which ships Prometheus v2.44.0, so pin that chart if you want the outputs to match and expect differences on a current chart. Beginners should start with book 2 or 5 first, because this one assumes you already read PromQL comfortably. About $47 when we checked. Check the current price on Amazon.
5. Hands-On Monitoring and Alerting with Prometheus
Muhammad Badawy’s book from AVA, Orange Education’s tech imprint, came out in June 2025. That makes it the only publisher-backed Prometheus book here printed after Prometheus 3.0. It is also the most practical of the three: the contents run from Prometheus architecture and metric types to exporters for Linux, Windows, Docker and databases, then service discovery, labeling and relabeling, PromQL and Alertmanager receivers such as Slack and email.
Relabeling gets its own chapter, and that is the part most beginners get wrong. A relabel_configs rule that drops or rewrites the wrong label changes which series you store without any error, so you only notice when a dashboard looks wrong weeks later. A book that walks through it with lab output beats a reference page for that reason. Pair it with our guide to monitor website uptime with the blackbox exporter for a second exporter to practice on.
Two reservations. The paperback runs 212 pages per Amazon’s catalog, so it is a guided course rather than a reference. And neither the publisher page nor anything else we found states which Prometheus release the labs target, so being newer doesn’t guarantee Prometheus 3 behavior. It’s the cheaper Prometheus pick at about $35 when we checked; the Kindle edition sits under a separate listing. Check the current price on Amazon.
6. Modern Network Observability
David Flores, Christian Adell and Josh VanDeraa wrote the only book here for network engineers, and it is also the most complete lab on the list. It collects metrics over SNMP and gNMI with Telegraf and logs with Logstash, enriches them from Nautobot as the source of truth, stores them in Prometheus and Loki, then builds Grafana dashboards, alerts and Prefect-driven automation on top. The final chapter applies machine learning to anomaly detection on the same data.
The lab repository pins everything, which makes it the easiest book here to reproduce exactly. Its Compose file defaults to Prometheus v2.52.0, Grafana 10.4.4, Loki 3.1.1 and Alertmanager v0.26.0, and the network side runs on containerlab with Arista cEOS routers. The authors document a setup on a DigitalOcean droplet, and it needs Docker 26.1.1 or newer and containerlab 0.54.2 or newer.
The catch is the cEOS image, which you download from arista.com with an Arista account before any lab starts. Server and application teams also get less from it than the title suggests, because the data comes from routers in a containerlab topology, not from web services. If metrics from switches are your job, it’s the best value on this page at $32 when we checked, a third under its list price. Check the current price on Amazon.
7. End-to-End Observability with Grafana
Ajay Reddy Yeruva and Vivek Basavegowda Ramu’s BPB title is the budget pick, spread across 20 chapters. The first half is a dashboarding course: the interface, panels, data sources, organizing dashboards and alerting. Chapters 3 and 5 teach the old Graph panel, an AngularJS panel that Grafana 12 force-migrates to Time series, so expect those screens to differ. The second half adds Grafana Cloud, AIOps monitoring and two chapters the other Grafana picks don’t have, on the blackbox exporter and synthetic monitoring. It closes with dashboards for a performance-testing workflow. That reflects Basavegowda Ramu’s background as a performance architect.
It is the oldest Grafana book on this list, from July 2023, and the contents page has no Tempo or OpenTelemetry chapter, so it won’t teach you traces. Treat it as a dashboard and alerting course with a strong performance-testing angle rather than a full observability book. Performance testers who need to watch a load test in Grafana get the most from it. About $25 when we checked. Check the current price on Amazon.
Adjust book examples for Prometheus 3 and Grafana 13
Every book above that pins a version pins Prometheus 2.x or Grafana 10, so some examples break on current releases. Most of the breakage is predictable, and both projects list it: the Prometheus 3 migration guide and Grafana’s what’s new and breaking-change notes. These are the changes you are most likely to hit while following a book:
| What the book shows | What current releases do | What to change |
|---|---|---|
holt_winters() in PromQL | Renamed to double_exponential_smoothing() and gated behind a feature flag in Prometheus 3 | Rename it and start Prometheus with --enable-feature=promql-experimental-functions |
A [5m] range sometimes returning 6 samples | Range selectors are left-open, so evenly spaced data returns 5 | Widen short subqueries such as [1m:1m] to [2m:1m] before using rate() |
--enable-feature=agent or remote-write-receiver | Feature flags removed in Prometheus 3 | Use --agent or --web.enable-remote-write-receiver |
| Targets with no Content-Type header scrape fine | Prometheus 3 fails the scrape | Fix the exporter or set fallback_scrape_protocol |
| Metric names limited to letters, digits and underscores | UTF-8 metric and label names accepted | Set metric_name_validation_scheme: legacy if you want the old rules |
| Legacy dashboard alerts on graph panels | Legacy alerting removed in Grafana 11 | Build the alert in Grafana Alerting instead |
| AngularJS panels and data source plugins | Off by default in Grafana 11, removed in Grafana 12 | Use the React version of the panel or plugin |
grafana-cli plugins install and grafana-server | Both commands removed in Grafana 13 | Use grafana cli plugins install and grafana server |
| Promtail shipping logs to Loki | Promtail reached end of life on March 2, 2026 | Move to Grafana Alloy; alloy convert --source-format=promtail translates the config |
Two more Prometheus 3 changes catch fewer readers but are harder to debug. The . in PromQL and relabeling regular expressions now matches newlines, so a .* matcher can select more series than it did on 2.x. Remote write also stopped defaulting to HTTP/2, which matters if a book’s Mimir or Thanos example relies on it. The simplest way to learn from these books is to run the lab on the version it pins, get it working, then repeat the step on a current server built with our Prometheus and Grafana on Kubernetes guide and fix what the table above predicts. If the work you’re preparing for is running these systems on call, our SRE books list covers the practices around them.
