Fleet health
Routers with a fresh sampleCoverage counts routers that reported within three collection intervals. A router without a fresh sample is shown as not reporting, not as down.
Coverage counts routers that reported within three collection intervals. A router without a fresh sample is shown as not reporting, not as down.
Configure checks, follow latency and loss, and monitor from your local sites.
Loading quality monitoring…| Device | CPU | Status |
|---|
Loading…
Loading…
Devices each router sees directly on its ports (MNDP, LLDP and CDP). Matched links are drawn on the map automatically.
| Device | Connects to | Seen by | Check |
|---|---|---|---|
| Loading… | |||
| Neighbour | Address | Seen by |
|---|---|---|
| Loading… | ||
Only the last 30 minutes of polls are used, so an unplugged link disappears on its own. Add an unmanaged neighbour from Routers & Sites to monitor it.
The score counts systems with no flags (including partial Docker-only checks) out of all systems. A partial check is not proof of safety.
| System | Type | OS / PHP | Containers | Disk used | Health | Scanned |
|---|
| Target | Type | Access |
|---|
See the routers, interfaces and alarms that need your attention.
Actual samples from the past 24 hours.
Latest known collector status.
Follow a link over time, including its most recent measured rates.
RX and TX rates are calculated from consecutive device samples.
| Router / site | Interface | RX | TX | Link speed | Errors Δ | State | Sample |
|---|---|---|---|---|---|---|---|
| Waiting for interface data. | |||||||
| Device | Site | Role | CPU | RAM | Loss | Latency | MTU | Sessions | Status |
|---|
| Device | Host | Site | Role | Vendor | Last Seen | Status | Actions |
|---|---|---|---|---|---|---|---|
| No live devices onboarded yet. | |||||||
| OLT | Site | Vendor | ONUs | Attention | Ping | Status | Actions |
|---|
Review current collector freshness, resources and routing checks.
Connect your network monitoring before running an assessment.
Get an evidence-linked review from a configured Ollama model.
No AI review has been generated.
The AI receives a bounded set of pseudonymous monitoring observations, not customer records, device names, addresses, credentials or router configurations. Review every suggestion against the source data.
View labels configured in Splynx for customer accounts.
| Label ID | Customer label |
|---|---|
| No labels loaded. | |
In Settings → Splynx, use your Splynx server's base URL. Your API key needs permission to view customer labels.
See known provider labels and fully unpaid invoices past their due date.
| Customer | Provider label | Earliest due found | Unpaid due months | Blocked since |
|---|---|---|---|---|
| No customer data loaded. | ||||
A blocked account and unpaid invoices do not prove why it was blocked. Months count distinct due-date months for invoices marked not_paid; partially paid invoices are excluded. Actual blocked duration needs verified status history.
Tickets currently identified as open, new, in progress or waiting.
| Ticket | Subject | Status | Assigned ID | Created |
|---|---|---|---|---|
| No ticket data loaded. | ||||
SLA timing, paused time, business hours and reopened tickets require your actual Splynx fields and support rules before a reliable breach can be shown.
Compare equipment recorded on completed jobs against stock issues and returns, including serial numbers and the person who booked it out.
A discrepancy needs both sources and is a review item, not proof of missing stock.
Queue missed calls and calls answered by each team member.
| Queue / team member | Answered | Missed | Abandoned |
|---|---|---|---|
| No call report loaded. | |||
Queue missed calls and individual missed calls measure different things. Calls outside queues are not included. Yeastar may require different API versions for older call records; its report date format must match your PBX.
| Site | Equipment | How it's running | Units | Cancelled (90d) | |
|---|---|---|---|---|---|
| Loading… | |||||
Every unit marked cancelled, with when, who cancelled and the reason. Cancellations picked up from Splynx have no reason until someone adds it.
| When | Unit | Customer | Who cancelled | Reason |
|---|---|---|---|---|
| Loading… | ||||
| Submitted | Customer | Last day | Reason | Splynx |
|---|---|---|---|---|
| Loading… | ||||
Installations customers booked on the website, from yesterday on. Each one has a Splynx ticket for the install team. Confirm a booking to put it on the Splynx calendar and email the customer; change the date when the provider gives another day.
| Day | Customer | Place | Notes | Splynx |
|---|---|---|---|---|
| Loading… | ||||
Faults customers logged on the website. When their site already had an outage or maintenance, they were told and no ticket was opened unless they asked for one.
| Reported | Customer | Place | Problem | Outcome |
|---|---|---|---|---|
| Loading… | ||||
People who checked coverage and signed up, or asked to hear when fibre reaches them. Each one becomes a Splynx lead with a ticket for sales.
| Submitted | Person | Place | Package | Splynx |
|---|---|---|---|---|
| Loading… | ||||
Routerboards, OLTs, switches, radios, power and anything else installed here. Link an item to a monitored router or OLT to see whether it is online.
| Equipment | Make / model | IP / MAC | Status |
|---|
Customers who gave notice, from the website form or an end date set in Splynx, and the day each unit goes empty. Striped units on the map below.
| Goes empty | Unit | Customer | Reason | From |
|---|
Click a unit to change its status, link it to a Splynx customer or see its history. Units linked to Splynx follow their Splynx status automatically.
| Target | Type | Latency | Jitter | Loss | Status |
|---|
Each check sends five pings; one lost packet equals 20% loss. Two consecutive bad readings create an incident.
| Check / target | Measured from | Latency | Loss | Last sample | State | Actions |
|---|---|---|---|---|---|---|
| Loading checks… | ||||||
Recorded round-trip time and packet loss for the selected check.
A Raspberry Pi at each site polls its routers, switches and OLTs, runs ping checks and finds everything the site can reach. It keeps collecting when the link to the NOC is down and uploads the backlog afterwards.
| Collector / site | Connection | Collects | Uploads | Computer | Version | Actions |
|---|---|---|---|---|---|---|
| Loading collectors… | ||||||
| Severity | Alarm & evidence | Router / site | When | Status | Actions |
|---|
| Circuit / site | Capacity | RX / TX | Packets/s RX / TX | Utilisation | Last sample | Status | Actions |
|---|
Burstable-billing rate for a circuit: 5-minute averages with the top 5% removed. Compare it with your transit or backhaul commit.
| Device | Reached through (upstream) | Behind it | Now |
|---|---|---|---|
| Loading… | |||
Alarms behind a down device are still recorded and shown in Incidents, labelled with the upstream device. Their phone alerts are held. If a device is still down after its upstream device recovers, its alert is sent then. Recovery messages are never held.
Minutes devices were unreachable each day, added up. Hover a day for its uptime.
| Site | Uptime | Outages | Downtime |
|---|---|---|---|
| Loading… | |||
Worst first. Each square is one day: green meets the target, amber is close, red is below, grey has no data.
| Device | Uptime | Daily | Outages | Downtime | Data |
|---|---|---|---|---|---|
| Loading… | |||||
Every period a device stopped answering polls, newest first. Kept for two years, long after raw samples are pruned.
| Device | Started | Ended | Lasted | SLA |
|---|---|---|---|---|
| Loading… | ||||
Solid lines are RX, dashed lines are TX. Rates come from consecutive counter reads; a counter reset shows as a gap.
| Interface | RX now | TX now | Packets/s RX / TX | Link |
|---|
Read-only. Each live view stops after 15 minutes; a router can have two open at once. Live readings are not stored; history comes from the regular collector.
| Window | Applies to | Time | State | Actions |
|---|---|---|---|---|
| Loading… | ||||
If a problem is still open when a window ends, its held alert is sent then. Recoveries during the window are not sent. Ending or cancelling a window is recorded in the operator audit log.
Loading…
| Incident | Area | Started | State | Actions |
|---|---|---|---|---|
| Loading… | ||||
Automatic incidents go up when a router suddenly loses PPPoE sessions and come down by themselves when the sessions are back. Customers see the title, area and message only, never router names or session counts.
Loading…
| Offline device | Site field | On status page | Fix |
|---|---|---|---|
| Loading… | |||
A device is on the status page only when its Site matches a live site under Sites. Complexes, estates and businesses are shown by name; POPs, towers and data centres are all shown as one “Part of our network” row. A router that says online but has had no reading for 10 minutes counts as offline.
Two readings in a row are needed to open or clear an alarm, and "not charging" needs the bank low on mains for the whole window set under Settings, Alarm thresholds. A missing reading never clears an alarm, but one that has not been seen again for hours while readings keep coming is cleared. Mains off is a warning; it turns critical, and reaches phones, when the battery is about to run out. Runtime left is a rough estimate from the last 30 minutes of voltage, and lead-acid voltage falls faster near empty.
Every router that reports its power supplies. One PSU failing while another is OK means the router lost redundancy (a warning). If a router only has one feed by design, list it under Settings, Alarm thresholds, Single power feed routers.
| Site | Router | Power supplies | Read |
|---|---|---|---|
| Loading… | |||
Bar height is the cycle time as a share of the poll interval. Red cycles ran longer than the interval; amber cycles had a failed router.
| Device | Completeness | Failed polls | Saved by retry | Gaps | Last good poll | Last error |
|---|---|---|---|---|---|---|
| Loading… | ||||||
Each site collector polls its own devices. While its link is down it keeps the readings and uploads them when the link returns, so they still count towards completeness.
| Collector | Connection | Devices | Waiting to upload | Computer |
|---|---|---|---|---|
| Loading… | ||||
| Channel | Sends | Last result | Status | Actions |
|---|---|---|---|---|
| Loading… | ||||
A summary of the last 24 hours with charts: bandwidth, alarms, outages, PPPoE, OLTs and ONUs, site power and internet latency. It goes to every channel with “Send the morning network report” ticked.
Choose how serious each kind of alarm is and what happens when it fires. “Call me” sends it to every channel, including PagerDuty phone calls, whatever its severity. “Message only” never phones. Changes apply the next time the alarm fires.
| Alarm | Last 30 days | How serious | What happens |
|---|---|---|---|
| Loading… | |||
The last 50 alarms and recoveries across all channels, newest first.
| When | Channel | Alarm | Result |
|---|---|---|---|
| Loading… | |||
| User | Role | Two-factor | Last sign-in | Status | Actions |
|---|---|---|---|---|---|
| Loading… | |||||
Settings, devices, circuits, maintenance, alarm acknowledgements and user changes, newest first. The full change log with before and after for every change, and changes waiting for approval, are on Security → Change control.
| When | Who | What | Details |
|---|---|---|---|
| Loading… | |||
Each MikroTik is read (read-only) and every change to its configuration is kept here line by line: accounts, services, firewall, NAT, routes, addresses, VLANs, PPP, DHCP, queues, scripts, the scheduler, BGP and more. Passwords and keys are never stored.
| When | Router | What changed | Who was signed in | |
|---|---|---|---|---|
| Loading… | ||||
Accepting a change applies it exactly as it was asked for, in the name of the person who asked. You cannot accept your own.
Loading…
Every change these accounts make waits for approval. Check with the person first. Unfreezing asks for your own password, so a stolen session alone cannot undo it.
Red items leave a door open. Fix them first.
Loading…
Entries cannot be edited or deleted, and each one is sealed with the one before it.
| When | Who | What | Result | Details |
|---|---|---|---|---|
| Loading… | ||||
| Check | Certificate | Issuer / names | Domain registration | Checked | Actions |
|---|---|---|---|---|---|
| Loading… | |||||
Fix the cause first (usually an infected customer device or an open mail relay), then ask the list for removal. Some lists clear by themselves once the abuse stops.
| Address | List | Why | Since | Remove |
|---|---|---|---|---|
| Loading… | ||||
Router addresses and public PPPoE/DHCP pools are found on each MikroTik automatically. Add other ranges you own, such as the prefixes you announce over BGP.
| Watch | Range | State | Checked | Actions |
|---|---|---|---|---|
| Loading… | ||||
Checks every list straight away without saving anything, for example a customer's address after they say a website blocks them.
Spamhaus, Barracuda and SpamCop listings are critical alarms because most mail servers use them; the others are warnings.
| List | What it means | Alarm | Links |
|---|
| Address | List | Listed | Cleared |
|---|
| Device / site | Sensor | Reading | Observed | Actions |
|---|
| Router / session | Remote ASN | State | Uptime | Prefixes | Observed |
|---|
Loads counters on demand from a configured FastNetMon HTTPS API. Values are source counters; validate their units, sampling and coverage with your detector.
| ASN | Reported bytes | Reported packets | Reported flows |
|---|
A router sends syslog to a collector on its private network. The collector forwards supported events outbound to NOC OS; router HTTPS is optional. Map the router’s exact outgoing IP below.
| Router / site | Collector | Router source IP | Collector status | Last event received | Mapping |
|---|---|---|---|---|---|
| Loading collector mappings… | |||||
UDP syslog can be lost or spoofed. Restrict the collector listener to the private network and trusted router IPs. Silence does not confirm that logging is working. Remapping or disabling a source can reject older queued events.
Choose the accounts and source networks expected on each router. These lists classify new observations; they do not change router access rules.
| Router / site | Policy | Approved accounts | Approved networks | |
|---|---|---|---|---|
| Loading policies… | ||||
Receives supported security logs through a site collector, or through optional RouterOS HTTPS polling.
| First seen by NOC | Router / site | Change | Router account | Source / method | Policy / feed |
|---|---|---|---|---|---|
| No configuration observations loaded. | |||||
Only supported formats are shown. The router account does not prove who used it or whether a change was authorised. Source and method are shown only when present; command text is not stored.
Receives supported security logs through a site collector, or through optional RouterOS HTTPS polling.
| Observed | Router | Account | Source | Method | Event | Account policy | Source policy | Feed |
|---|---|---|---|---|---|---|---|---|
| No login observations loaded. | ||||||||
| Received | Target | Reported action | Severity | Incoming traffic | Incoming PPS | Source |
|---|---|---|---|---|---|---|
| Open this page to load reports. | ||||||
Optional direct RouterOS HTTPS is needed for a complete visible account snapshot. Local syslog can report supported account changes without HTTPS, but cannot list all existing accounts. The first snapshot is a baseline to review.
| Router / site | Visible accounts | Last check | Review |
|---|
| Observed | Router | Account | Change | Before → after |
|---|
Optional direct RouterOS HTTPS snapshot. This is not needed for local syslog collection. Short sessions can start and end between checks; the monitoring account may appear.
| Router / site | Account | Source | Method | Policy | Snapshot received |
|---|---|---|---|---|---|
| No sessions loaded. | |||||
Read a named RouterOS address list on demand. Being on a list does not prove that an address is currently blocked.
| Address | Comment | Disabled | Dynamic | Timeout reported |
|---|---|---|---|---|
| Choose a router and enter a list name. | ||||
| Device | Captured | Source | Hash | Change |
|---|
Every site, unit, fibre layout, alarm history and setting lives in the NOC database. It is backed up every night and test-restored once a week.
Loading…