Ask which signal is missing

An offline device and a stopped service answer different questions. Offline means the monitoring system has stopped receiving expected check-ins. A stopped-service observation means a collector obtained a particular service state. Neither finding, by itself, explains the cause.

Start with the timestamp, source and freshness of the evidence. A service shown as Running yesterday tells you nothing reliable about its state after today's connectivity outage. Treat a missing observation as missing, not healthy or failed.

Use a simple decision table

Device check-in Service observation First useful step
Fresh Fresh Running Check application-level symptoms if users still report a fault
Fresh Fresh Stopped / Failed Inspect the named service and its dependencies locally
Fresh Unknown / unavailable Check name, query support and permissions
Stale Last known Running Investigate the missing device check-ins first
Never received None Verify enrollment and initial startup

These are investigation paths, not automated diagnoses. A process can be Running while its application is unresponsive. A monitoring agent can stop reporting while other applications continue working.

If check-ins stopped

Confirm whether the device is awake, powered on and reachable through your normal administration method. Laptops can sleep; virtual machines can be paused; network paths can fail. A ping timeout alone does not establish a powered-off device because ICMP may be filtered.

Check the local monitoring service, then its last accepted server response. On an enrolled Ubuntu device:

systemctl is-active fleet-monitor-agent
sudo fleet-monitor-agent status

On Windows:

Get-Service -Name FleetMonitorAgent

The Windows command reports the local service state. Use the companion or portal to verify server confirmation. These commands do not repair enrollment. Investigate the configured HTTPS address, certificate validation, clock and network before replacing the identity.

If only the selected service stopped

Inspect the exact name and observation timestamp. Determine whether it should be continuously running. Look for a recent maintenance window, deployment or configuration change. Check the service's own logs and dependencies through an authorized local administration path.

For a systemd service, a query can separate load and active state:

systemctl show ssh.service --property=LoadState,ActiveState,SubState

Replace the example unit with the service under investigation. systemctl and Microsoft Get-Service document the local inspection commands. A query failure is evidence about the query, not proof that the service stopped.

If many devices go offline together

Compare transition times before treating each device as a separate hardware failure. A shared network path, application outage or deployment could explain a cluster of missing check-ins. Verify the monitoring system through an independent operational check. A dashboard cannot guarantee its own availability simply by displaying healthy devices.

Keep original identities and incident timestamps while investigating. Re-enrolling several devices can obscure continuity and create more work without fixing the shared cause.

Verify recovery explicitly

Look for a new accepted device check-in and fresh service state. Historical graphs and a Running process are useful context, but neither substitutes for end-to-end confirmation. Record what changed and whether the evidence supports the diagnosis.

IronAgents keeps offline incidents separate from service incidents. Unknown or stale service readings do not falsely recover an active incident. The troubleshooting documentation gives the corresponding product checks.