Skip to content

What's New

This page is a tour of what GridNMS does today, grouped by capability rather than by release. GridNMS Cloud picks up new capabilities automatically. Self-hosted instances are told when a new release is available and apply it when you choose. Collectors and log forwarders can update themselves once you turn that on for them; anything installed before that was available takes one reinstall first. See Updates for how delivery works.

Model a service — “Customer Portal,” “Internet Access at the Denver office” — as a graph of the components it’s actually built from: managed devices, discovered applications, plain endpoints, or active checks, wired together with hard and soft dependencies. GridNMS rolls health up the graph for you, so a degraded cache reads as degraded, not a false down, and a failed database takes the service with it. Every check runs from one of your own collectors, so internal services — an intranet app, an internal DNS server — are covered right alongside internet-facing ones. See Service Monitoring.

  • Checks survive a collector outage. Service checks follow the same collector assignment rules your devices do, so if the collector running a check goes offline, GridNMS re-homes the check onto another healthy collector that can reach the target. If none can, the check reads unknown rather than falsely reporting the service down.
  • A latency chart per check, in the service graph. Expand any check inside a service’s dependency graph to see its own response-time history, so you can tell an outright failure from a gradual slowdown without leaving the view.
  • Service-level objectives. Give a service an availability target over a rolling window and its graph shows attainment, error budget remaining, and how fast that budget is burning right now. A built-in burn-rate detection can alert you while there’s still budget left to protect.
  • Click a red segment to find out why. Any red or amber slice of a service’s status timeline opens an evidence drawer naming the failing monitor, its target, the reason it reported, the response time, the collector that observed it, the events raised in that window, and the individual check results behind the failure — with a shareable link to the same view.
  • Alerts name the cause, not just the detection. A service alert’s subject and body carry the service, component, impact and the specific failure — for example Website — DOWN (Web frontend: HTTPS 502 Bad Gateway) — and grouped Cases carry the same detail.
  • The graph flags what happened in the window. A component that went down and recovered inside the range you’re viewing is marked as such, instead of looking identical to one that was healthy the whole time.
  • Ready-made telemetry detections. GridNMS ships a library of built-in, vendor-neutral detections — CPU and memory utilization, temperature, UPS battery, interface errors and discards, latency degradation, and unexpected device reboots. They work the same way on every supported vendor, arrive disabled by default (no day-one alert storm), and you enable just the ones you want under Configure → Logs and Events → Detections. Each threshold detection clears itself when the reading recovers.
  • Detections update themselves. New built-in detections are delivered automatically as they’re released — like the vendor and service-name reference data, with version tracking. Your own edits, enable/disable choices, and deletions are always preserved.
  • Interface thresholds. Set traffic and utilization thresholds directly on an interface. Breaches raise events through the same detection pipeline as everything else, and recovery is recorded when the value comes back down.
  • Container thresholds. Alert on a container’s CPU, memory, or network and disk I/O right from its detail drawer on the Containers page. See Container Monitoring.
  • Alert when a machine stops reporting. A scheduled rule can now watch the host metrics a log forwarder collects from its own machine — CPU, memory, disk, network — and fire when they go quiet, which is how you catch a server that stopped reporting rather than one that reported something alarming. Write it with domain = "host"; GridNMS matches the machine whether it reports a short or a fully qualified name, and attaches the event to the device that forwarder is linked to. See Writing rules.
  • Durable collector offline alerting. GridNMS reliably alerts you when a collector goes offline and clears the alert when it reconnects, so a site that drops off is never missed.
  • Maintenance windows. Schedule planned downtime for devices or sites so expected outages don’t generate noise. Notifications are suppressed for the window, and upcoming maintenance is shown on the dashboard and in reports.
  • Silence a flapping device without losing its events. For a device that blips but is otherwise fine, turn off Reachability Alerts on its Overview tab. Up/down events are still recorded and still group into cases; only the email or page stops, and only for reachability — a real fault on that same device still reaches you. GridNMS records who turned it off and why. See Notifications.
  • Cases are the alert surface. Related events are grouped into a single case, and the case — not the individual event — is what notifies you. Case detail includes a timeline viewer for the case’s member events (expandable rows with a field summary) and a panel explaining why the case was created. See Cases.
  • Case rules with priority. Write your own grouping rules for how events roll up into cases. When more than one rule could match, the highest-priority rule wins — one case, no duplicates.
  • A permanent event record. Events are immutable, append-only records of what happened — they’re never edited or overwritten. Investigation status (open, closed, ownership) lives on the case instead. The events list shows which case each event belongs to, with a Hide case-managed events toggle to focus on what isn’t already being handled.
  • Clearer events list. Events show the detection that produced them, how many log lines matched, and a richer description — with a one-click view matching logs link to jump from an event to the underlying logs.
  • Notifications link straight to the source. Email and webhook notifications deep-link to the exact event or case they’re about.
  • Log management and search. Collect syslog and SNMP traps from your devices, search across all of them, and drill into the fields that matter with a live fields sidebar. Traps carrying binary and text payloads are decoded correctly, and forwarded syslog carries the full structured record — not just the message text — so more fields are searchable.
  • Container and service logs land already sorted. Logs shipped from a host’s journal arrive under the container or system service that produced them — for example docker:my-app or linux:sshd — instead of one undifferentiated bucket, so you can search one service without writing a parser first. Every parser GridNMS applies is listed on the Parsers page.
  • Configurable retention. Choose how long to keep logs — by age or by a fixed storage size — so your history stays predictable.
  • Device configuration backup & diff. GridNMS backs up device configurations on a schedule, keeps a version history, detects real changes (ignoring volatile lines like timestamps), and shows side-by-side diffs. Each device gets a Config tab with versions, changes, the full config, and download. See Configuration backups.

  • Automatic device identity. Make, model, software version, operating system, and DNS name are detected automatically and shown (and editable) on the device overview. A daily system scan keeps interfaces and identity details current, including forward and reverse DNS cross-checks. When GridNMS knows an address’s hardware (MAC) address, the device overview shows it beside that address.

  • A device that blocks ping still reads Up. Plenty of hosts drop ICMP on purpose. If the reachability check fails but a service monitor aimed at that same device has answered in the last few minutes, GridNMS shows the device as Up and takes its Last seen time from that check, rather than calling a machine that is plainly answering you Down. A monitor reporting a service down never marks the device down — that can mean only that service stopped. See Devices.

  • API-based monitoring — Cisco Meraki. Monitor cloud-managed devices through the vendor’s API, alongside your SNMP and SSH devices. Manage API credentials under Configure → API Credentials and attach an API target per device or per device class.

  • Wider device reach. Devices whose SSH login only supports interactive prompts (common on appliances) can be polled, and per-device SSH credential overrides are honored even when they differ from the device’s class.

  • GridNMS can create the SSH key for you. Rather than generating a key pair yourself and pasting the private half in, select Generate a key for me on a device’s SSH credentials. GridNMS keeps the private key in its encrypted store — it’s never displayed — and hands you the public key to install on the device. See Devices.

  • Ready-made scripts for the read-only login. The documentation now carries a reviewed script for Linux, macOS, and Windows that creates the dedicated unprivileged account GridNMS logs in as, installs the public key with key-level restrictions, and grants no elevated rights. Run one on the host you want monitored instead of assembling the steps by hand. See Devices.

  • More from an agentless server. A Linux or macOS server monitored over SSH now also reports network throughput per interface and identifies its own make and model; Linux adds per-disk read and write throughput. No agent, still just a read-only login. See Server Monitoring.

  • On-demand collection. Run a single data collector on demand from a device’s detail page, instead of re-running the whole pack, with clear reporting if it fails.

  • Scan one host on demand. Discovery can check a single address instead of a whole network — useful for confirming one new host before you onboard it. GridNMS still verifies the address sits inside a network the collector is allowed to reach. See Discovery.

  • Host names from DNS, per network and off by default. Looking up a name for each responsive address is now its own switch — DNS lookups for hosts, on each network under Configure → Sites & Networks — rather than something port scanning decided for you. Turn it on for the networks where you want names and leave it off elsewhere, without changing whether or how often those networks are scanned. See Hosts.

  • Port scanning without committing to a schedule. Allowing a network to be port-scanned and scanning it on a timer are now two switches rather than one. Turn on Enable port scanning on this network to sweep it when you choose, and leave Scheduled scans off to keep it from running on its own. Turning the permission off turns the schedule off with it, so switching it back on later does not quietly restart recurring scans. Networks already scanning automatically keep their interval. See Networks & Sites.

  • Sort the device list by any column. Click Name, IP Address, Class, Collector, Last Seen, or Status to order the whole inventory by it — not just the page on screen — and click again to reverse. Addresses sort numerically rather than as text, Status brings the devices that need attention to the top, and the order travels in the page address, so a refresh or a shared link opens the same view. See Devices.

  • Vendor monitoring packs. Ready-to-use monitoring packs for Cisco (IOS-XE and Nexus), Juniper, Fortinet, Aruba, APC/UPS, Arista, HPE/Aruba, QNAP, and generic SNMP hosts — so common gear is monitored correctly out of the box.
  • MIB browser. Browse the MIBs behind your monitoring packs directly in the app, grouped by the pack that provides them.
  • Clearer service and topology views. Discovered services are labelled with their well-known names (for example ssh (22/tcp)), with duplicates de-duplicated for a cleaner picture.
  • Smarter metric collection. Polling intervals are tuned per data point so fast-changing counters stay responsive without over-collecting slower metrics.
  • Load-balancing profiles with a standby option. Choose how devices are assigned across the collectors at a site — including a dedicated backup collector that stays idle and only takes over when the primary collector goes offline.
  • Decide which release your organization runs. Configure → Service Management → Software lists the Collector and the Log Forwarder for each operating system and shows, per row, the release your agents are following, the newest one published, and when each became available. Activate a new release when you’re ready, or set how soon it is taken automatically — immediately, or after one, three, or five days — separately for each operating system, so you can move Linux ahead of macOS. Every row starts on Manual only, nothing is activated for you until you choose it, and Roll back returns an operating system to the release it was on before. A new release only reaches your agents once it has been activated here.
  • Collectors that update themselves. Turn on automatic upgrades for a collector and it carries every part of itself forward on its own — the collector, its updater, and the small programs beside it — each one checked against our signature before anything is swapped, and rolled back if the new version doesn’t come up. Upgrade now takes the latest release on demand, without waiting, and release groups let you take a new version on a few collectors first and the rest once it has settled. A collector installed before this was available needs one reinstall from the current download, and the Software section on its Settings tab says so rather than offering a switch that couldn’t work.
  • See whether a collector’s data is actually arriving. A collector keeps two separate connections to GridNMS — one for management, one for the data it sends — and a healthy management connection says nothing about the second. The Collectors list now reads both, so a collector whose data has stopped reaching GridNMS shows No data arriving instead of a green tick, and its Overview tab reports Data delivery on its own line with the address to check. Where delivery can’t be confirmed either way, it says so rather than claiming a good state.
  • A collector’s own outage no longer looks like your whole site failing. When a collector loses its path to the internet, every device and service check it reaches that way fails at the same moment — through no fault of theirs. Before reporting anything down, a collector now checks whether it can still reach the outside world, and holds back only the results its own outage can account for: devices and checks it reaches across the internet hold their last known state, while a switch or a server on its own network is still reported down as usual. A service check against a public address reads unknown rather than down, and a target that actually answered — a refused connection, or a name that doesn’t resolve — still reports down, so a genuinely broken service still alerts. A long outage stops being held back after about an hour, so nothing stays quiet indefinitely. See Collector health.
  • Choose where an agent installs. Pass --install-dir when you install a collector or a log forwarder to keep it on another volume, with everything it needs underneath that one directory — on Windows the log forwarder takes -InstallDir. Re-running the installer keeps your location, and asking to move an existing install is refused rather than quietly leaving the agent unable to find its enrollment.
  • Collectors that work behind TLS inspection. A network that inspects outbound encrypted traffic used to need an inspection exception for a collector to deliver anything. Turn on Work with TLS inspection on the collector’s Settings tab instead: it identifies itself on every request and delivers over a publicly trusted connection your inspection device can read, with no exception to arrange.
  • Log forwarders that update themselves. Turn on Auto-upgrade for a Linux or macOS forwarder and it installs new releases through its collector — downloading in the background, restarting once, and rolling back on its own if the new version doesn’t start. Upgrade now takes the latest release on demand, without waiting. A Release group on each forwarder sets how long after a release becomes available it takes it automatically, so a new version can prove itself on a few forwarders before the rest follow.
  • Fleet-wide forwarder policies. Set automatic updates on for every newly approved forwarder, push that setting across the whole fleet in one action, and generate a shared install token so new hosts are approved the moment they connect instead of queuing for a decision.
  • Forwarder hosts identify themselves. A connected forwarder reports its manufacturer, model, and operating system name and version automatically — no module to turn on — so a host’s entry in your inventory is filled in without you typing it.
  • NVIDIA GPU monitoring. Turn on the GPU Metrics module for a Linux host’s forwarder profile and every NVIDIA GPU in that host reports its utilization, memory, temperature, power draw, clock speeds, fan speed, and ECC error counts — with a card and charts per GPU.
  • GPUs on the device, and alerts per card. A device whose host reports GPUs gets its own GPU tab, appears in Device Metrics and Metric Explorer, and can be thresholded on any GPU reading. Each card is judged on its own: one hot GPU in a four-card host raises one alert naming that card, and cooling down clears only that one. A Service Monitor component can watch GPUs the same way.
  • Windows Event Log collection. Pick the Event Log channels a Windows host should send — Security and System with one action, or any channel you name.
  • Full configuration export & import. Export your entire GridNMS configuration (devices, credentials, monitoring, users and roles, notifications, detections, dashboards) as a single password-protected, encrypted file. Import previews every difference — row by row, field by field, classified New / Changed / Unchanged — and applies only what you tick.
  • API tokens. Create personal access tokens under Profile → API Tokens to script against GridNMS. A token can carry your full user permissions or a narrower scope you choose at creation — read-only, a specific set of permissions, or limited to particular sites — so an integration only gets the access it actually needs. A published API reference documents the endpoints.
  • Data feeds. Curated reference datasets — device vendor identification and well-known service names — update automatically over the collector’s existing connection, with per-dataset status, version history, and the ability to refresh or roll back to a specific version (Configure → Service Management → Data Feeds).
  • Local overrides. Override any reference entry with your own value under Configure → System Configuration → Reference Data — your overrides survive dataset updates.
  • Site-aware roles and permissions. Roles can be scoped down to individual sites, so a user can be given access to only the sites they’re responsible for. Every view, list, and action respects each user’s authorized sites, and you can switch your active site on the fly.
  • Mandatory multi-factor authentication. Sign-in is protected with a passkey as your second sign-in step, an authenticator-app fallback, and recovery codes, with policy controls for your organization.
  • Hardened collector connections. Per-site IP allowlisting and one-time diagnostic access tokens give you tighter control over what each collector can reach.
  • Single sign-on decides group membership. Write rules that read an attribute your identity provider sends at sign-in — a group claim, a department, a job code — and place the user in the matching GridNMS group automatically, re-checked every time they sign in. A group an administrator set by hand is never overwritten. See SSO Mapping.
  • Scheduled log forwarder credential rotation. Replace each forwarder’s ingest credential on a schedule instead of by hand. Rotation runs one forwarder at a time and the old credential keeps working until the host confirms the new one, so nothing stops delivering. See Managing Log Forwarders.
  • Update notice with one-click apply. Self-hosted administrators see an in-app banner when a new server release is available and can apply it with a single confirmed click — the update health-checks itself and rolls back automatically if something goes wrong. See Updates.
  • Self-serve signup and customer portal. Sign up, manage your subscription, and handle billing from a self-service portal.
  • About dialogs everywhere. Every GridNMS screen can tell you exactly which version you’re running, with a copy-to-clipboard summary for support requests.
  • Consistent device search. Anywhere the app asks you to pick a device, you get the same fast, searchable device picker.
  • Shareable views, and Back that works. What you’re looking at lives in the page address across the app — the query you built in Graph Explorer, the search and device you drilled into in Metric Explorer, the log class you opened in the Data Model Explorer, the reference dataset you were editing. Bookmark a view or paste the link to a colleague and they see the same thing, not an empty starting point; Back, Forward, and refresh behave the way you’d expect. See Graph Explorer, Metric Explorer, and The Log Data Model.
  • Richer dashboard panels for services and collectors. The Service Status panel adds a 24-hour timeline per service, and the Collector Status panel shows a row per collector with its heartbeat age, owned devices, and CPU, memory, and log-volume sparklines. See The Dashboard.

docs built 2026-11-12 · 945a2dec2