Skip to content
Sencai.docs

Monitoring

Infrastructure → Monitoring (/gravity/monitoring) is the live-health view for everything Sencai tracks on your behalf: your provisioned cloud instances and any servers running the fleet agent. It’s the input the rest of the Operations loop builds on - an incident starts from an alert raised here.

What’s monitored

Section titled “What’s monitored”
  • Cloud instances - CPU, memory, and disk utilization, network in/out, and load average, collected from each provisioned instance roughly once a minute.
  • Fleet agents - a table of every enrolled agent with its hostname, operating system, connection status, and time since its last heartbeat. An agent that hasn’t reported in over five minutes is flagged as stale.

Metrics only exist for resources Sencai actually manages. A cloud instance you haven’t provisioned through Sencai, or a server without the fleet agent installed, won’t appear here - see Provision & manage instances and the fleet agent guide to get something onto this page.

Reading the metric cards

Section titled “Reading the metric cards”

Each provisioned instance gets a card with its current CPU, memory, and disk usage. Cards color their metrics amber and red at fixed thresholds, so you can scan the page for trouble without reading every number:

MetricWarning (amber)Critical (red)
CPU70%90%
Memory75%90%
Disk80%95%

An instance the platform couldn’t reach for its last collection cycle is marked Error instead of showing stale numbers.

The Refresh button forces an immediate collection cycle instead of waiting for the next automatic one. Auto-refresh re-polls the page itself every 30 seconds so newly collected data appears without a manual reload.

Alert thresholds and firing alerts

Section titled “Alert thresholds and firing alerts”

Behind the metric cards, alert rules define a metric, a comparison (greater than, less than, and so on), a threshold, and a severity (info, warning, critical) - the logic that decides when a metric crossing a line becomes a firing alert rather than just a number on a card. Each rule also has a cooldown, so a metric that keeps flapping around a threshold doesn’t fire repeatedly within a short window.

Firing and resolved alerts appear in the Recent Alert Events list at the bottom of this page, each with its severity, the metric and threshold it crossed, and a timestamp. This list is the primary place to see alert activity today - how (or whether) a firing alert additionally reaches you outside the app is covered in Notifications & escalation.

What you need to configure

Section titled “What you need to configure”
  • Provision at least one cloud instance, or enroll at least one fleet agent, or there’s nothing for this page to show.
  • Nothing else is required to start collecting - metrics begin appearing automatically once a resource exists.

What’s next

Section titled “What’s next”