DocumentaciónHelp

How to read the risk

What the risk levels, health, unusualness, anomaly mode and signals shown by Cerebro mean.

On this page

Cerebro expresses a component's failure risk in several ways. This page explains each one.

Levels

Level

What it means

High risk

The failure probability exceeded your company's threshold. It usually opens a predictive alert.

Watch

The risk went up, but it hasn't reached high risk yet. It's worth keeping an eye on.

Normal

Low risk.

No data

There is no prediction yet.

Cerebro learns the High risk threshold for each component based on your company's alert budget. See Settings.

Risk at 1, 6 and 24 hours

It is the probability, as a percentage, that the component will fail within that time frame. For example, 6 h risk: 34 %. Very small values are shown with decimals, such as 0,4 %.

Health

Health is 1 minus the 24-hour risk: a healthy component is close to 1.00 and moves away from 1 as the risk rises.

Anomaly mode and unusualness

To estimate a percentage, Cerebro needs to have seen real failures. Until it has enough, it works in Anomaly mode: it compares how the component behaves against its own history and shows:

  • Usual behavior or Unusual behavior, instead of a level with a percentage.

  • The Unusualness vs. its history, as a percentage.

  • Disk about to fill up, if the disk forecast sees that it will fill up soon.

In this mode Cerebro doesn't generate alerts based on unusualness alone. You'll see a notice like “Learning: N more failures needed for the supervised model”, with how many failures there are and how many are needed.

Signals

Next to the risk you'll see the main signals:

  • ▲ raises the risk, for example a metric that moves away from the usual.

  • ▼ lowers the risk.

A signal marked as third-party data comes from logs or from the server agent.

Signals accompany the risk; they are not a diagnosis of the cause.

Trend

On the pages you'll see whether the risk rose, fell or is stable over the last 6 hours, compared with the previous 6 hours. In anomaly mode, the same applies to unusualness.

Related

Search Argunis Monitoring

Type at least 2 characters to search Argunis Monitoring.

↑↓ to navigate · Enter to open · Esc to close