The Opservo Health Standard · OHS v1.0
One health number you can actually argue with
Most tools hand you a single opaque “health” score you can’t question. OHS is the opposite: a deterministic 0–100 rating where every point lost traces to a named rule, its threshold, the current value — and the fix.
In short
The Opservo Health Standard (OHS) is a deterministic, open-rule method for scoring any Linux server from 0 to 100. Seven weighted categories (Performance, Security, Storage, Reliability, Maintenance, Network, Configuration) are scored independently and blended. Every lost point names the rule, the metric, the threshold, the current value and the fix, and each finding carries a confidence level (Definitive, Likely or Heuristic). The same facts always produce the same score, and the standard is versioned.
crm-prod · health
Every one of the 18 lost points is a rule you can open, understand, and fix.
Why it’s different
A score is only useful if you can trust it
A black-box number tells you something’s off but not what, why, or whether it even matters. OHS is built on four principles so the score earns its place on the dashboard — and so an AI recommendation on top of it can be trusted.
Deterministic
The same facts always produce the same score. No averaging of arbitrary signals, no hidden ML, no time-of-day drift. A server that hasn’t changed doesn’t wobble between 78 and 85.
Explainable
Every point lost is attributable to a named rule — with its ID, threshold, the current value, and the fix. If a server scores 82, you can see the exact 18 points and why.
Confidence-scored
Each finding is Definitive, Likely, or Heuristic — so a “firewall is disabled” fact is weighted differently from a “possible memory leak” hunch, and you can trust each one proportionally.
Extensible & versioned
New checks are rules against a fixed frame — dormant ones activate the moment their collector ships, with no change to the math. The standard is versioned (OHS v1.0) so scores stay comparable.
How the score is built
Facts in. Explanations out.
The score is never computed directly from a collector. Facts flow through rules and categories, so every number is traceable end-to-end.
Collectors
The agent reports facts only — disk 96% full, firewall off, one CPU sample at 97%. Collectors never compute a score.
Rules
Each rule turns a fact into a penalty: a metric, a threshold, a severity, a confidence and a fixed point cost.
Categories
Each category starts at 100 and subtracts the points of every rule that fired. It can never go below 0 or above 100.
Overall score
The seven category scores are blended by fixed weights into one 0–100 number — and a ranked list of what to fix.
Seven categories
What the score is made of
Each category is scored independently, then blended by a fixed weight. You see the category scores too — so you know whether a problem is performance, security, or storage at a glance.
Performance
weight 25%
CPU, memory pressure, load, context switches, CPU throttling
Security
weight 20%
Exposed ports, SSH hardening, firewall, brute-force guard, auto-updates
Storage
weight 15%
Disk & inode usage, read-only filesystem, SMART health, SSD wear
Reliability
weight 15%
OOM kills, kernel panics, segfaults, a served app returning 5xx
Maintenance
weight 10%
Pending & broken packages, orphaned packages, log-rotation health
Network
weight 10%
Gateway reachability, packet loss, DNS resolution, NIC errors
Configuration
weight 5%
NTP time-sync, swap configured, uncommitted config drift
A category whose collector hasn’t shipped yet scores a neutral 100 and is flagged low-confidence — a missing signal never silently drags a server down.
Security is one of the seven categories — see how we secure your servers and account →
Confidence scoring
Not every check is equally certain
OHS attaches a confidence to every finding, so the UI and AI recommendations can be trusted in proportion to how sure the signal is.
“Firewall is disabled.” “Disk is 96% full.” Acted on immediately.
A single CPU reading at 97% — a strong but single-sample signal.
“Possible memory leak,” “suspicious process.” Presented as worth checking.
Rating bands
From Perfect to Critical
The 0–100 number maps to a plain-English rating, so anyone on the team knows what it means without learning the scale.
See your servers’ health scores in 5 minutes
One command to install. Every score comes with the exact reasons behind it — and the fix.
FAQ
Health score questions
How do I run a server health check?＋
By hand, on a Linux server: check load and CPU with uptime and top, memory with free -h, disk space and inodes with df -h and df -i, failed services with systemctl --failed, recent errors with journalctl -p err -b, and listening ports with ss -tulpn. Opservo runs those checks, plus security, maintenance and configuration rules, on every heartbeat and turns them into the 0–100 OHS score, with the rule, the current value and the fix behind every point lost.
What is a good server health score?＋
On the Opservo Health Standard, 90–100 is Very Good to Perfect, 80–89 is Good, 70–79 is Fair, 60–69 Needs Attention, 40–59 is Poor and 0–39 is Critical, which usually means a finding such as a disk about to fill, a failing disk or an exposed database port. Because every lost point is a named rule, the number matters less than the list beneath it.
How is the 0–100 score calculated?＋
Each server is checked against a fixed set of rules grouped into seven categories: Performance, Security, Storage, Reliability, Maintenance, Network and Configuration. Each rule has a threshold and a point penalty; the category scores are blended with fixed weights into one number. There is no machine learning in the score, so the same facts always give the same result.
Why does Opservo show a rule ID like STOR_DISK_001?＋
So the score can be argued with. The ID identifies the exact rule, its threshold and its weight; the page shows the current value that triggered it and the command that fixes it. You can dismiss a finding you disagree with, and the standard is versioned (OHS v1.0) so scores stay comparable over time.
Does the health score include security?＋
Yes. Security is one of the seven categories and covers exposed database ports, root SSH login, password-only SSH, missing firewall, failed-login bursts, unscoped passwordless sudo and pending security updates. The same rules feed the separate security posture view, so there is one scorer, not two.