> ## Documentation Index
> Fetch the complete documentation index at: https://docs.rubixkube.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# How to monitor infrastructure health with RubixKube

> Set up continuous health monitoring across Kubernetes, AWS, GCP, and VMs. Learn what each dashboard surface tells you and when to act.

This tutorial walks through the monitoring workflow: connect an environment, read the dashboard, spot anomalies early, and set the handful of controls that keep signal-to-noise high. Takes about fifteen minutes from a fresh workspace.

## Prerequisites

<CardGroup cols={2}>
  <Card title="A RubixKube account" icon="user" href="https://console.rubixkube.ai">
    Free tier works for this tutorial.
  </Card>

  <Card title="At least one environment connected" icon="plug" href="/getting-started/connect-your-environment">
    Kubernetes, AWS, GCP, or a Linux VM. Mix of any of them is fine.
  </Card>
</CardGroup>

## Step 1: Read the Dashboard

The Dashboard is your daily check-in. Four tiles tell you whether the system is worth investigating right now.

| Tile                 | What it means                                    | When to pay attention                 |
| -------------------- | ------------------------------------------------ | ------------------------------------- |
| System Health        | Blended score across every connected environment | Below 95% or trending down            |
| Active Insights      | Count of anomalies RubixKube is watching         | Any new insight since your last visit |
| Intelligent Analysis | RCA reports ready to read                        | New report since last session         |
| Agents               | Health of the observer and cloud-side agents     | Anything other than all green         |

<Tip>
  Open the Dashboard once a day even when nothing is on fire. Most anomalies show up as a subtle drift, not an outage.
</Tip>

## Step 2: Check your topology

Open **Infrastructure Topology**. You should see every resource the Observer has discovered, grouped by environment.

* **Green edges**, dependencies behaving as expected.
* **Yellow edges**, degraded signals (higher latency, elevated error rate, resource pressure).
* **Red edges**, active incidents.

Clicking any node shows the last hour of signals for that resource plus any insights currently attached to it. Good starting point for deep dives.

## Step 3: Tune Insights to your team

Open **Magic Insights**. Each insight is an anomaly the system thinks a human should know about.

<Steps titleSize="h3">
  <Step title="Filter to your services">
    Use the environment and namespace filters to narrow to what your team owns. Bookmark the view.
  </Step>

  <Step title="Set the severity threshold">
    Start at Medium. Too many low-severity cards train people to ignore the list.
  </Step>

  <Step title="Subscribe to the ones that matter">
    Each insight has a Follow action. Followed insights post to your notification channel when their status changes.
  </Step>
</Steps>

## Step 4: Connect a notification channel

Health monitoring is only useful if the right person sees the signal. Connect a channel you already live in.

<CardGroup cols={2}>
  <Card title="Slack" icon="slack" href="/using/integrations">
    Channel-level routing for insights and RCAs.
  </Card>

  <Card title="Microsoft Teams" icon="microsoft" href="/using/integrations">
    Team-channel delivery for insights and RCAs.
  </Card>

  <Card title="PagerDuty" icon="bell" href="/using/integrations">
    Promote critical insights into on-call pages.
  </Card>

  <Card title="Linear" icon="list-check" href="/using/integrations">
    Turn RCAs into tickets with one click.
  </Card>
</CardGroup>

## Step 5: Ask Chat a monitoring question

Chat is the fastest way to pull a specific view without learning a query language. A few prompts worth bookmarking:

```
What changed in the payments service in the last hour?
Which resources have the highest error rate today?
Any hosts above 80% memory right now?
Show me deployments that rolled back in the last 24 hours.
```

Answers come with cited evidence, so you can jump from the reply straight to the underlying events.

## What healthy monitoring looks like after a week

<CardGroup cols={2}>
  <Card title="System Health is stable" icon="heart-pulse">
    Sits between 95 and 100% most of the time. Dips correlate to known events.
  </Card>

  <Card title="Insights have owners" icon="user-check">
    Your team either acts, dismisses, or routes every new insight. Few stale cards older than a day.
  </Card>

  <Card title="Topology reflects reality" icon="diagram-project">
    Newly deployed services appear. Decommissioned resources drop out within the hour.
  </Card>

  <Card title="Chat answers with evidence" icon="comments">
    Your team uses Chat instead of grepping logs for quick questions.
  </Card>
</CardGroup>

## Common questions

<AccordionGroup>
  <Accordion title="How fast does RubixKube detect a new issue?">
    Most anomalies surface within one to two minutes of the underlying signal. The OPEL loop runs continuously rather than on a fixed cron schedule.
  </Accordion>

  <Accordion title="What is the difference between an Insight and an RCA Report?">
    An Insight is an anomaly worth attention. An RCA Report is a full causal chain with evidence and recommended fixes. Not every insight becomes an RCA, only the ones that look like they have a single identifiable root cause.
  </Accordion>

  <Accordion title="Can I monitor multiple environments at once?">
    Yes. Every connected environment feeds into the same Dashboard, Insights list, and knowledge graph. Use the environment filter to narrow to a single one when needed.
  </Accordion>

  <Accordion title="What if I only want alerts, not a dashboard?">
    Connect Slack or Teams, filter Insights to the severity you care about, and subscribe. The dashboard becomes optional.
  </Accordion>
</AccordionGroup>

## Related guides

<CardGroup cols={2}>
  <Card title="How to Automate Incident Remediation" icon="wand-magic-sparkles" href="/tutorials/automate-incident-remediation">
    The next step after monitoring: when an insight becomes an incident.
  </Card>

  <Card title="Talk to your infra" icon="comments" href="/tutorials/talk-to-infra">
    Go deeper with Chat for on-demand investigations.
  </Card>
</CardGroup>