Uptime monitoring from outside your own infrastructure that alerts you the minute your site goes down — upordown.online
In a large company, an outage usually ends with a document. There's a template, a review meeting, action items with owners and a tracker that nags people until they're done.
In a team of two or three developers, an outage usually ends with someone saying "OK, it's back," a sigh of relief and a return to whatever was interrupted. Maybe a message in chat: "it was the disk again." Then, four months later, it's the disk again, and nobody can remember what they did last time.
I think small teams need written postmortems more than big ones, not less. Here's why, and what a small-team version looks like.
Your memory is the only incident database you have
Big companies have many layers of institutional memory: incident trackers, on-call handbooks, people whose job is reliability. A small team has whatever is in the heads of the people who were there.
That memory is surprisingly bad. Within a few weeks, details blur. What was the error? Which command fixed it? Did we fix the cause, or just restart the thing? Was it the same problem as March, or a different one that looked similar? Without a written record, every repeat outage starts from zero.
And small teams change. Someone leaves, someone joins, someone goes on holiday during the next incident. The one person who remembers how to fix it is the one who isn't there.
"We don't have time" is the argument for doing it
The usual objection is time. A small team is busy shipping, and writing documents feels like overhead.
But consider the cost on the other side. An outage you've seen before, without notes, means rediscovering the diagnosis under pressure, often at a bad hour, while customers wait. Twenty minutes of writing after the first incident can save hours during the second.
And the time doesn't need to be large. The value of a postmortem isn't in its length; it's in existing at all.
What a small-team postmortem looks like
Forget the twelve-section enterprise template. For a small team, five headings are plenty:
# 2026-11-03: Checkout returned errors for 40 minutes
## What happened
Short description in plain words, written so a new teammate understands it.
## Timeline
- 14:05 First alert (checkout synthetic check failing)
- 14:12 Started investigating
- 14:30 Found the cause
- 14:45 Fixed and confirmed
## Why it happened
The real cause, and the reasons it wasn't caught earlier.
## What we did to fix it
The exact commands, settings or changes. Copy-paste them.
## What we'll change
One to three concrete actions, each with a name next to it.
Write it within a day or two, while details are fresh. Keep it in your repository, in a postmortems/ folder, so it's searchable and lives next to the code.
That's it. No meeting required, though a short call to read it together is worthwhile.
The next postmortem is easier if you already have an outside timeline of when the site went down and came back. upordown.online keeps that record, and alerts you at the start of the incident.
The rules that matter
A few principles make even a short postmortem useful:
Blameless, genuinely. In a team of two, blame is personal, and it kills honesty quickly. Write about what the system allowed to happen, not who made a mistake. "The deploy script didn't check disk space" rather than "Alex forgot to check disk space."
Include the commands. The most valuable part of an old postmortem is often the exact fix. "Cleared the logs" is far less helpful than the actual command you ran and the path.
Few actions, actually done. Two actions completed beats ten actions forgotten. If an action isn't going to happen, don't write it down.
Note what went well. If an alert fired quickly, or a runbook worked, say so. It tells you what to keep.
Link to it. Reference the postmortem in the fix's commit or pull request, and in any related runbook. A document nobody can find is a document nobody reads.
The real payoff: patterns
After a year, a small team with written postmortems has something remarkable: a list of every significant outage, why it happened and what was done. Read through it and patterns appear.
Maybe half your outages trace back to one fragile dependency. Maybe they cluster around deploys, or around a particular time of month. Maybe alerts were consistently slow, or consistently fired for the wrong thing. These patterns are almost impossible to see from memory, and obvious on paper.
That list is also a powerful tool when you need to argue for time to fix something. "This has caused four outages this year, here are the write-ups" is far more persuasive than "I think this is flaky."
When to write one
Not every blip needs a document. A sensible rule for a small team:
- Always: anything customers noticed, anything that lost or corrupted data, anything involving security.
- Usually: anything that took longer than half an hour to fix, or woke someone up.
- Sometimes: a near-miss that could easily have been serious.
If you're unsure, write a short one. The worst outcome is a few paragraphs nobody needed.
Start with the last outage
If your team has never written a postmortem, don't wait for the next incident. Write one for the most recent outage you can remember, using the five headings above. You'll be surprised how much you've already forgotten, which is exactly the point.
Knowing when something broke and how long it lasted, which is the start of every timeline, is what we help with at upordown.online.
Related reading
- Postmortem (hypothetical): our health check was so thorough it took every instance out of the load balancer at once
- Our monitor said 100% uptime while the CDN served a 6-hour-old homepage
- Case study (hypothetical): the site that was down only for people on IPv6, and every monitor said it was fine
- Rotating an API key or database password? A 12-point checklist so the rotation isn't the outage
Does your team write anything down after an outage, and if you looked for notes on the last one, would you find them?
Top comments (0)