How it works
How Sentry talks to your infrastructure
Each method answers a different question. Sentry actively tests some services, queries some devices and listens to records others send. It then merges findings for the same target and time window.
Three interaction modes
-
01
Tests
Sentry initiates the request and records response time, content and result.
-
02
Queries
Reads counters and state from devices at set intervals.
-
03
Listens
Receives event records sent by devices, parses and orders them in time.
Message flow
- Sentry test client
- Switch / AP
- RADIUS
- EAPOL-Start The test client starts authentication.
- Access-Request The access device forwards the request to RADIUS.
- Access-Challenge The server asks for credentials; step time is measured.
- Access-Accept / Reject Result, VLAN assignment and total time are recorded.
Illustrative flow · not live data
Tests · Port / transport: EAP over LAN · RADIUS UDP 1812/1813
Can a device join the network with authentication, and how long does it take?
Evidence collected
- Per-step authentication time
- Accept / reject result
- Assigned VLAN
- Timeout point
How AI reads this data
Looks at which step accumulates delay; compares with RADIUS records and port state to rank possible causes.
Boundary: Runs with a defined test identity; does not touch real user sessions.
Message flow
- Sentry
- Switch / router / server
- GET ifOperStatus Port operational state is queried.
- up / down Port state is stored with time.
- GET ifInErrors Error counters are read at intervals.
- Counter value The delta from the previous read is calculated.
Illustrative flow · not live data
Queries · Port / transport: UDP 161 · traps UDP 162
What do the device's interfaces, counters and hardware state say?
Evidence collected
- Port up/down state
- Traffic and error counters
- CPU, memory, temperature
- Device uptime
How AI reads this data
Matches counter growth with events on the same port; separates physical, cabling and load-related possibilities.
Boundary: Read-only; device configuration is never changed.
Message flow
- Network device / service
- Sentry
- %LINK-3-UPDOWN The device sends a port event.
- radiusd: timeout A service timeout record arrives.
- auth failure An authentication failure record arrives.
- Parsing Records are grouped by source, severity and time.
Illustrative flow · not live data
Listens · Port / transport: UDP / TCP 514
How did devices and services record what happened at that moment?
Evidence collected
- Source and severity
- Event text
- Timestamp
- Repeat frequency
How AI reads this data
Groups similar records, reduces noise and places them in the same window as active test results.
Boundary: Sentry never writes back to senders; it only listens.
Message flow
- Sentry
- Server:443 / 5432 / 1883
- SYN A connection attempt is made to the target port.
- SYN-ACK The port is listening; handshake time is measured.
- ACK The connection completes.
- FIN The connection is closed cleanly.
Illustrative flow · not live data
Tests · Port / transport: TCP · target service port
Even if the server is up, does the service accept connections?
Evidence collected
- Open / closed / filtered
- Connect time
- Timeout
- Consecutive failures
How AI reads this data
If ICMP succeeds but the port is closed, it treats the issue as closer to the service layer and suggests checking the service.
Boundary: Connects and closes; no data is sent to the service.
Other active checks
- ICMP
Is the target reachable, and what are latency and loss?
- HTTP / HTTPS
Does the app answer with the right status and time, is the certificate valid?
- MQTT
Does the broker accept connections, does the expected topic message arrive?
Findings merge into one picture
802.1X test latency, SNMP error counters on the switch port and RADIUS Syslog records land on one timeline. AI reads that picture and suggests possible causes and next checks; verification and decisions stay with your team.
Example scenario · not live data
The server is up. Why have orders stopped?
Three separate signals. One investigation story.
01 / 04
Signals arrive.
ICMP checks reachability. HTTP tests the service response. Syslog brings in the event recorded by the application.
Three methods, three different questions.
Reachable · no service response · error recorded
02 / 04
Findings join one incident.
Results from the same target and example time window come together while retaining their sources.
A reachable host does not invalidate a service failure.
Network access continues. The issue appears at the service layer.
03 / 04
AI focuses the investigation.
The reachable server, missing HTTP response and application record are read together. AI links possible explanations to evidence.
Possible cause: the application or a related service layer.
Start by reviewing service state and application records from the same window.
04 / 04
The team knows where to look.
Findings to verify and suggested checks come together in one summary. Intervention stays with the team.
Value: focused investigation instead of scattered alerts.
Review → verify evidence → let the team decide on intervention
Illustrative interpretation flow. No definitive root cause or automatic intervention is shown.
Demo request
Get in touch
Tell us about your environment and the targets you want to monitor; let's work out the right setup method and pricing together.
A monitoring plan that fits your needs
- 01Number of targets to monitor
- 02Check types and frequency
- 03Usage needs
Email: sales@mintyfi.com