You do realize fascism is Zeus punishing Prometheus for sneaking off with fire hidden inside a fennel stalk right? He sends Pandora as a gift to Epimetheus (Hindsight) who, despite Prometheus (Foresight) having warned him, oh yes plenty of warning, thoughtlessly accepts what was clearly a very bad no good offer.
So, we've been screwed for quite some time; ambition started it, denial assured it. As Hesiod once said: Thanks, Obama.
Thunder is back!
In this 16-minute episode, I talk with @saswatamcode about what Thanos is, what problems it solves, and how it extends Prometheus for large-scale Kubernetes monitoring.
(((o(ﾟ▽ﾟ)o)))
https://youtu.be/wsZAJGGXv4k
Телеметрия, диагностики и компилятор
В современном мире невозможно себе представить взрослое приложение, которое не экспортирует телеметрию. Метрики — важнейший атрибут поддерживаемого софта; для всех более-менее профессиональных технических специалистов термин « visibility » давно вытеснил прочие остальные баззворды наподобие « test coverage » и « continious integration ». Примерно двадцать лет назад я впервые узнал о философии отказоустойчивости Джо Армстронга, которая сводится к тому, что инструменты, помогающие разработчику не принести в код ошибку — штука хорошая, но по гамбургскому счету бесполезная, потому что не существует такой приблуды, которая сделает программиста идеальным. Разработчики, даже самые лучшие, будут ошибаться, и первоочередная цель инструментария — минимизировать потери от неизбежных ошибок, а не элиминировать оные полностью . Достижимые цели всегда лучше хрестоматийно правильных. Отсюда вырос знаменитый «слоган» эрланга « Let It Crash! », высмеиваемый и всегда неверно трактуемый теми, кто неутомимо отслеживает все исключительные ситуации… и спотыкается об отсутствие нужного для их корректной обработки контекста в месте ловли. « Let It Crash » — означает не « Хрен с ними, с ошибками », а « Случилась какая-то хрень в рантайме, но мы к ней готовы » . ¡NB! В тексте нет прямых примеров использования телеметрии в ООП/ФП проектах, но я глубоко убежден, что этот текст будет полезно прочитать, даже если вы просто перекладываете джейсоны из пустого в порожнее на шарпах, котлине или хаскеле. Telemetry it!
For the fun and to test if my #prometheus queries were correct, I started following the public timeline of mastodon_DOT_social from my #GoToSocial account. Guess what time it was where I did it...
Маршрутизация силами Haproxy, DoH, GeoIP, защита сервисов через mTLS и выгрузка метрик в Prometheus, настройка ACME.SH
Работаем с Haproxy, маршрутизация по GeoIP и ограничения, настройка mTLS для защиты сервисов, выгрузка метрик в Prometheus. Настройка панели 3X-UI для работы с Unix Socket и персональный DNS over HTTPS.
Как правильно использовать rate() в Grafana: от мониторинга до расчёта SLO
Функция rate() в PromQL необходима для вычисления средней скорости изменения метрики в секунду за определённый период времени. Она часто используется для мониторинга таких показателей, как: Как подружить rate() и Grafana
So after upgrading #Prometheus from 3.1.0 to 3.3.0 (nixpkgs), suddenly my Grafana dashboard is missing data for several panes related to network traffic and disk usage. I see the underlying metrics are still visible in the node exporter data, but not in Prometheus. Not seeing anything obvious in the change logs -- what am I missing?
My aim for the next day, was to setup a Dashboard for #crowdsec. There is a Web Application hosted by CrowdSec, but they have a limit of 500 Alerts per month. Since many IPs try to brute force my ssh login (I disabled password login, btw), I reach this quota in the first week of a month. Very sad.
So I did some research and figured out, the most generic solution is to use the metric Endpoint, which is compatible with #prometheus I also found an office dashboard from CrowdSec for #grafana!
So my goals were set. I started with Prometheus and collected the metrics (no big deal). Then I got my Grafana up and running and configured the data source for Prometheus. After that imported the Dashboard and I just worked!
Now I got hyped and also added my reverse proxy Traefik to the Prometheus jobs. There was a very up-to-date dashboard for Traefik which I could use. Now I can also display what services are used the most and see more insights of my home lab.
The rest of the day, I used to tweak my dashboards. Now I can see in a Donate Diagram the usage of my services :D See the screenshots below.
3/4
What a crazy weekend! I improved my home lab setup and got it on a new level. I want to share with you guys, how I got my Web Application Firewall (WAF) with CrowdSec, how long it took me to setup Prometheus and Grafana to visualize CrowdSec alters and my Traefik services. Last but not least, I will talk about my new Forgejo Runner (CI/CD) and how I got Docker In Docker working for Build Jobs. Below you find three toots for every topic.
#opensource #selfhosted #docker #crowdsec #traefik #waf #grafana #prometheus #forgejo
1/4
Several weeks ago I started to monitor my #selfhosted services with #prometheus. Two weeks ago I set alarms to notify me about possible problems. And today I made updates and buckups so I used alert silencers for the first time. It's so straitforward. I love it.
Next huge task I am plannig is to migrate it all from docker to kubernetes.
#GrafanaCON hands-on labs are almost sold out!
Create a custom data source plugin: 4 seats left
#Grafana as code: 4 seats left
#OpenTelemetry & #Prometheus native telemetry pipelines w/ Alloy: 4 seats left
Dashboard workshop: SOLD OUT
Register before we sell out: https://grafana.com/events/grafanacon/2025/hands-on-labs/
Counters, gauges, histograms, summaries!
The 4 essential metric types—explained clearly in under a minute.
From the first Thunder episode with Iris Dyrmishi
See the full episode:
https://youtu.be/1EtqYh5sTl8
I took the opportunity of the #GoToSocial upgrade to 0.19.0 to also start polling the #prometheus endpoint using the #VictoriaMetrics vmagent. This lead to identifying a bug (I think) so the gentle devs now have another PR to deal with 
Regarding OSes, this is #OpenBSD polling and storing metrics from #NetBSD and rendering them using #OmniOS and #pkgsrc.
Who said I was a sane person 
Something amazing of learning more about #Grafana and #Prometheus as part of building the @tuist server is that we’ll build on that knowledge to provide a /metrics for endpoint for organizations and developers, and pre-made Grafana dashboard if they want more flexibility over how they’d like to visualize and interact with their projects’ data.
Tuist plays the role of collecting, standardizing, and making that data accessible.
I replaced 10 MiB Go bloat with an AWK script and a tiny HTTP/CGI server. This time it’s the #Unbound metrics exporter. It does the some job as the unbound_exporter written in Golang, but in a fraction of total lines of code, disk space and memory usage. And it’s perhaps even safer, given how common vulnerabilities are in Golang’s core libraries…
https://github.com/jirutka/unbound-exporter-cgi
Many #Prometheus exporters (scrapers) can be rewritten in much efficient way, they are usually very simple. Another example is the Node Exporter – OpenWrt rewrote it in a few lines of Lua.
You may think that CGI is an ancient technology; I surely wouldn’t recommend it for running any public website, but I think that for this use case – metrics exporter only accessed from a private monitoring network – it’s an ideal solution.
What was monitoring like before Prometheus?
Iris Dyrmishi breaks it down in this clip from the first Thunder episode—a short recap of our full 
️ Enlightning session.
Full epidsode here:
youtu.be/1EtqYh5sTl8
arc) 