Skip to content

Monitoring & Managing Collectors

Collectors are the workhorses that poll your devices and receive their logs and traps. GridNMS gives you two views into their health: a per-collector view for detail, and a fleet-wide view for an at-a-glance picture across every collector and background service.

A collector installed from the current download can update itself once you turn that on for it, on its Settings tab under Software. For any other collector, download the newer build from the Customer Portal → Downloads and re-run the installer on that host — it’s safe to run on an already-installed collector and won’t lose its enrollment. See Installing a Collector → Updating.

Go to Configure → Collectors. The list shows every registered collector with an Online or Offline indicator and when it was last seen.

A collector open on the Collectors page, showing its detail tabs. Configure → Collectors. Each collector shows online/offline status; open one for its Overview, Health, Settings, and Security tabs.

A collector talks to GridNMS over two separate connections: one for management (its configuration and updates) and one for data (the logs, metrics and scan results it sends). The indicator reads both, and the Overview tab shows each one on its own line — Management and Data delivery.

The row says What it means
Online (green) Connected for management, and its data is reaching GridNMS.
Online · waiting for first data (amber) Connected, but nothing has arrived from it yet. Normal for the first few minutes after you install a collector.
Online · data not verified (amber) Connected, but GridNMS can’t confirm whether data is arriving — see below.
No data arriving (red) Connected for management, but its data isn’t getting through.
Management connection down (red) Its data is still arriving, but GridNMS can’t manage it — it won’t pick up configuration changes or updates.
Unhealthy / Config rejected (red) The collector is reporting a problem with itself, or it refused the configuration GridNMS sent. The Overview tab’s Agent config status gives the reason.
Offline (grey) The collector is off, or its site has lost its connection. This isn’t treated as a fault.

When a row says No data arriving, open the collector: the Data delivery line on the Overview tab says which of three situations you are in, because they point at different places.

Data delivery says What it means
Never delivered data The collector enrolled within the past week and nothing has reached GridNMS since. The path from the collector to GridNMS was never open — look at the firewall, the proxy, or anything that inspects encrypted traffic between them. The message names the address and port the collector must be able to reach.
No data since (a time) The collector was delivering and then stopped. Something that was working has changed — a network change, a certificate, a proxy rule.
No data in the last 7 days The collector enrolled more than a week ago and has nothing in that window. GridNMS can’t say whether it ever delivered, only that it hasn’t recently.

A collector at a site that goes offline on a schedule (a nightly shutdown, for example) is Offline, not No data arriving: when its management connection is down too, that is one expected state rather than a fault, and it clears by itself once the site is back and data resumes.

An amber row is not an error — it means GridNMS is declining to claim a collector is fine when it can’t tell. Data not verified has three causes, and the Overview tab’s Data delivery line says which: the check couldn’t run just now, no ingest address is configured for this collector, or the collector is on a version too old to report what GridNMS needs to check delivery. That last one clears when you update the collector, not by changing anything on your network — so if a collector has been amber since you installed it and it’s an older build, update it before investigating your firewall.

Open a collector to see its detail tabs:

  • Overview — connection info and the networks assigned to this collector.
  • Health — live status and metrics: what it’s running, plugin status, resource usage over time, and pipeline stats for log and metric delivery. See Reading a Collector’s Health Tab for what each card means.
  • Settings — listen ports, log ingestion, which plugins are enabled, and software/version info. This is also where you choose where the collector ships its logs.
  • Security — revoke the collector’s identity, manage its IP allowlist, and review recent security events for it.

Revoke collector identity on the Security tab is what to reach for if a collector’s host is lost or you believe it has been tampered with. Select Revoke now and the cut-off is immediate:

  • Its credentials are rotated, so it can no longer fetch its configuration or the device credentials it was holding, and a reconnecting session is refused.
  • Any session it has open is force-closed within a few seconds — the tab tells you when that has been confirmed.
  • It is dropped from the set of collector identities the log pipeline recognises.

Revoking is not a substitute for containing the machine itself. If you believe the host is compromised, isolate it on your network as well — revoking removes its access to GridNMS, not your attacker’s access to the host.

The collector stays in your list with its site and network assignments intact — revoking cuts the machine off, it doesn’t undo what you configured. To bring a host back, re-join it with a fresh token; it is issued new credentials rather than reusing anything the old host held.

Devices that collector was responsible for have no one to poll them until another collector covers their network, so plan the replacement alongside the revocation rather than after it.

A collector runs several workers, each shown as a chip on the Health tab:

Plugin What it does
Monitor Liveness checks — is each device up?
Poller Collects metrics and details from devices.
Syslog Receives syslog messages from devices.
SNMP Traps Receives SNMP traps from devices.
Service Monitoring Synthetic HTTP / DNS / SMTP / SMB checks.
Discovery Active network discovery scans (on-demand, not continuous).

Each chip always shows whether that worker is enabled or disabled for the collector (turned on/off under Settings). On collectors that report a live runtime status, the chip also turns green with a Running label when it’s confirmed working; a worker that isn’t currently doing anything (for example, the trap listener when nothing is sending traps) simply reads enabled without the green confirmation — that’s normal, not a problem.

Metric bars, last-seen, and stale detection

Section titled “Metric bars, last-seen, and stale detection”

The Health tab shows recent history for the collector’s resource usage — memory, CPU, and disk — and, when the collector reports it, tunnel latency and a devices-monitored trend.

A collector reports in with a heartbeat about every 30–60 seconds. If GridNMS hasn’t heard from it in a couple of minutes, the last known figures are shown with a stale metrics warning. Stale metrics on an otherwise-online collector usually mean it’s under heavy load or its connection is unstable.

For the big picture, go to Configure → Service Management. This page shows the health of the background services that keep monitoring running, alongside collector status across your whole fleet — all on one screen.

The Service Management page showing service health and collector sessions. Configure → Service Management gives an at-a-glance view of background services and connected collectors.

At the top you get summary figures, and below them:

  • Services — each core service with a Healthy / Degraded / Unknown badge. A green dot means it’s fine; red means it needs attention.
  • Connected collectors — every collector currently connected, with how long it’s been up. If a collector you expect is missing here, it isn’t connected.
  • Trend charts — recent history for key measures (memory, response time, event rate, and connected collectors) so you can see whether something is trending the wrong way rather than just its value right now.

Use Service Management as your daily glance, and drill into an individual collector under Configure → Collectors when something there looks off.

  • Every expected collector shows Online, and its Overview tab shows Data delivery as Arriving.
  • Every plugin you rely on shows enabled, and where a live status is reported, it’s Running rather than crashed.
  • Resource usage (memory, CPU, disk) is well below the host’s own limits.
  • On the Health tab’s pipeline stats, backlog / queue depth stays near zero and metric-write drops aren’t climbing (see Reading a Collector’s Health Tab).
  • No stale metrics warnings.

A collector is Online but shows No data arriving

Section titled “A collector is Online but shows No data arriving”

The collector can reach GridNMS for management but its data isn’t getting through.

  1. Open the collector’s Overview tab and read the message — it names the ingest address and port the collector must be able to reach.
  2. From the collector’s host, confirm an outbound connection to that address and port is allowed by your firewall or proxy. The management connection and the data connection go to different addresses, so one working does not mean the other does.
  3. Once the host can reach it, data starts arriving on its own; the row returns to a normal Online within a few minutes.
  1. Confirm the collector process is actually running where you deployed it.
  2. Check its network path to GridNMS — the collector makes an outbound connection, so a firewall or proxy change can cut it off.
  3. Once connectivity returns, the collector reconnects on its own — you don’t need to re-register it.
  4. Check last seen on the Collectors list to gauge how long it’s been gone.

Devices stopped updating, but no down alerts arrived

Section titled “Devices stopped updating, but no down alerts arrived”

This is usually deliberate. Before a collector reports a device down, it checks whether it can still reach the internet — a handful of well-known internet addresses and a DNS lookup. If none of those answer, the collector concludes the problem is on its own side and holds back the device reports that losing its path would explain, rather than alerting on everything at once.

Which devices this covers depends on what the collector lost. Losing your internet connection says nothing about a device on the same local network, so the collector does not treat the two the same:

What the collector has lost Devices on your local network Devices it reaches across the internet
The internet, while its own network connection is up Reported down as usual Hold their last known state
Its own network connection Hold their last known state Hold their last known state

So a printer or a switch that is genuinely down is still reported during an internet outage, and only the devices the outage can actually account for go quiet. “On your local network” here means an address the collector reaches without leaving its own link — including a device with a public address on that link, which is normal on an IPv6 network.

While a device’s reports are held back:

  • It holds its last known state instead of flipping to down, so the Devices list keeps showing what was true before the collector lost its path.
  • A device that does answer is still reported up as normal.
  • Only the affected collector is involved — devices watched by your other collectors are unaffected.

It takes a couple of consecutive failed checks — around two minutes — before a collector reaches that conclusion, so a brief blip never triggers it. An internet outage that simply keeps going stops being held back after about an hour: from then on those devices report down, so a long outage never stays quiet indefinitely.

There is no separate alert for this, which is why a site that goes quiet can look like nothing happened at all. If a group of devices stops updating together:

  1. Check whether they are all watched by the same collector. If they are, that collector’s own connectivity is the thing to look at, not the devices.
  2. Open that collector’s row and read last seen. A collector that has also lost its management path shows as Offline.
  3. Confirm the collector’s host can still reach the internet. If your firewall blocks all outbound ICMP and DNS from that host by design, the collector cannot tell a real outage from its normal state, so this judgement is unreliable there — it has never seen the internet answer, so it will not claim to have lost it. Allow those checks if you want it to work at that site.
  4. Once the collector’s path returns, normal reporting resumes on its own and the devices update on their next check.
  1. Open the collector’s Health tab and watch the backlog / queue depth pipeline stat.
  2. If it drains back toward zero, it was a temporary traffic burst — nothing to do.
  3. If it keeps climbing, the collector is receiving more than it can ship. The collector may be undersized for its log volume, or its connection to GridNMS may be constrained.
  4. You can always confirm delivery directly in Logs & Log Search — see Reading a Collector’s Health Tab for what each pipeline stat means.
  1. Give it a moment — a crashed worker usually restarts automatically.
  2. If it doesn’t recover, the collector is likely under stress or hitting a limit. Check its resource usage on the Health tab.
  3. If it persists, see Troubleshooting.

docs built 2026-11-12 · 945a2dec2