If you’ve ever SSHed into a server at 2 a.m., run htop and df -h, and squinted at a log file hoping for a clue, this post is for you.
At Softuvo, we build and host software for about 30 companies. That adds up to 40+ production Linux servers running Laravel, WordPress, Node, Docker, nginx, MySQL and Redis. We don’t have a dedicated SRE. We have developers who do ops in their spare time.
Knowing a server is in trouble was never the hard part. The hard part was understanding why, and what to do next.
The tools didn’t fit
We tried what everyone tries. Uptime checks told us a site was down, and nothing more. Grafana dashboards looked great, but nobody opened them until after the outage. Datadog assumed we had a team whose full-time job was observability.
Every tool could tell us the CPU was high. None told us that nginx had stopped rotating logs and the disk would fill in six days.
Two real failures from September
Here’s what pushed us to write this up:
A failing disk nobody heard about. An office workstation’s disk failed and the filesystem went read-only just after midnight. Opservo, the tool we built, flagged it within a minute. But that workspace had no notification channel set up, so the alerts sat unseen. The detection worked and the delivery didn’t.
Eight weeks of silent alerts, and this one was on us. Our own Discord webhook had been rejecting every alert since late July, including through an eight-day Redis outage. Every alert was generated and then dropped. We only caught it on September 28.
We’re sharing the second one because “alerts that silently fail” is exactly the problem we set out to solve, and we hit it ourselves. Every channel now records the result of its last delivery attempt.
What Opservo does
One command per server. About 30 seconds later it has inventoried your services, containers, domains, ports and log sources.
A health score from 0 to 100 within five minutes. Every lost point names the rule, the threshold, the current value and the fix.
Explanations in plain English. You get what’s wrong, what changed and what to do, not just a graph.
Approve fixes from the dashboard, or let a policy apply known ones.
Safe by default. The agent is read-only, opens no inbound ports, uses signed outbound HTTPS, and logs every command in an audit trail. Remote actions are opt-in per server.
Who it’s for
Agencies, small managed-service providers and small SaaS teams running roughly 5 to 50 Linux servers with nobody watching them full-time. If you already have an SRE team running Datadog well, you don’t need it. If your developers do ops on the side, you probably do.
It’s free for two servers, with no card required.
Read the full story: Why We Built Opservo: 40 Production Servers, 30 Companies, no SRE
Try it on the server that scared you last: getopservo.com
Top comments (0)