---
title: "Seventeen minutes on 8 June 2026"
summary: "The one outage we know of since we started running client sites: what failed, why it took everything with it, how we noticed, and what changed since."
author: "Samuel Krauss"
author_title: "Founder"
publisher: "Elchi Studios"
published: 2026-10-12
updated: 2026-10-12
url: https://elchi.dev/en/journal/seventeen-minutes-on-8-june-2026
language: en
tags: ["incident","operations"]
words: 443
---

# Seventeen minutes on 8 June 2026

*By Samuel Krauss, Founder. Published 12 October 2026.*

> The one outage we know of since we started running client sites: what failed, why it took everything with it, how we noticed, and what changed since.

On the morning of 8 June 2026, at about ten o'clock, everything we ran was unreachable for about seventeen minutes: our clients' websites, our own site and our services. It is the one outage we know of since we started running websites for clients in December 2024. This is the report we owe for it, written later than it should have been.

## What happened

Our code used a third-party library for licence checks. That library ran into a fault of its own, and the way our code called it meant that its failure was not contained: it took the whole system down with it. Nothing else was wrong. No server failed, no database was damaged, no data was lost. Once the cause was found and the dependency on the library was taken out of the path that serves pages, everything came back.

We do not name the library. The fault was in how we used it as much as in the library itself, and naming it would put the blame somewhere it only half belongs.

## How we noticed

The honest answer: because everything was suddenly gone. At that time no monitor watched our services from outside. We found out the way a visitor would have, which is the wrong way round. The seventeen minutes are our own reconstruction of that morning, not a measurement. That is also why we do not state an exact availability figure for the time before: there was nothing that measured it.

## What changed

Three things.

- **The library is gone.** It is no longer part of any system we run.
- **A dependency that is not needed to show a page can no longer stop a page from being shown.** A licence check, a statistics call, a notification: if one of them fails, the page still answers and the failure is logged. This is now a rule we check in review, not a hope.
- **Four independent monitors watch our services.** They run in different places and alert separately, so a single failing monitor cannot hide an outage. Since our own cluster went live on 3 October 2026, [status.elchi.dev](https://status.elchi.dev) publishes what is measured, including every disruption with its real length.

## Why we write this

A company that tells you its numbers should also tell you about the time the numbers were bad. If seventeen minutes on a Monday morning affected you and you never heard from us about it, this is the explanation, and we are sorry it comes this late.

From now on, every disruption that a client could notice gets a report like this one in the journal, with what failed, how long it lasted and what we changed.