On-Site Agent update (optional): Upstream Path Discovery on a Site served by an On-Site Agent on Linux (every On-Gateway Agent included) could come back with most traces as "No reply" and every hop Unreachable; the v2.11.3 Agent fixes it (see Network Performance below). Skipped v2.11.2 or v2.11.1? The same upgrade brings the fix for false SNMP rate spikes on a Site whose SNMP an On-Site Agent collects, Live Traffic's live flows and Live DNS on the On-Gateway Agent, plus its Tailscale paths. Open Settings - Multi-Site, expand your Site, press Upgrade, and run the command for your install (on an On-Gateway Agent, Run It for Me runs it for you).
An Upstream Path Discovery that proposes far fewer wrong hops (and no longer comes back "No reply" on Sites served by an On-Site Agent), cleaner SNMP interface rates, a Transit RTT that one noisy router no longer drags around, and LAN Speed Test fixes for On-Site Agents. See the v2.11.2 release notes, the v2.11.1 release notes, and the v2.11.0 release notes for what's new in v2.11.0+.
Monitoring
Network Performance
- Fix: Upstream Path Discovery came back mostly "No reply" - on a Site served by an On-Site Agent on Linux, with every hop Unreachable. It needs the v2.11.3 Agent.
- Fix: ISP RTT near 0 ms, or ISP Health's Packet Loss and ISP Network scores pulled down by the wrong hops (thanks @micallan17 for the report) - Upstream Path Discovery proposed the container's own bridge as an ISP hop on a Docker install with bridge networking, and the whole path (CDN, DNS, and AWS endpoints included) on a run whose network lookups failed. If your ISP RTT reads near 0 ms, or Access ISP hops lists CDN or DNS hosts, re-run discovery and commit: the wrong Targets are paused or moved to the right type, keeping their history.
- Upstream Path Discovery proposes fewer wrong hops - your ISP's gateway in IP passthrough or router mode comes up unchecked as Customer premises instead of as an ISP hop (re-run and commit to pause one you already monitor). A transit router far slower than the rest (200+ ms among routers at 20 to 30 ms) comes up unchecked too, and an access hop you uncheck, or pause in Latency Targets, stays that way on the next run.
- Fix: a transit router your traffic never passes through proposed as transit - an exchange peering address on the path to a DNS or CDN endpoint can answer through another carrier when probed directly, and the old check could let one through. It's left out as Different route now; re-run and commit to pause one you already monitor.
- Also fixed: a Target discovered on a WAN whose discovery you later removed listing under your primary WAN unlabeled (Latency Targets now shows a badge naming that WAN); and a reload after picking LAN as the scope opening an empty chart.
ISP Health
- Fix: Path Localization placing a problem its traces never saw - a destination whose pings lost packets while every trace to it ran clean still got a verdict placing the problem past the last clean router, or on the return path. It now says the traces ran clean and can't place the problem. On new episodes, the hop table also leads a degraded destination's row with its own ping loss and RTT, where it showed only the trace's (0% loss while the pings lost several percent).
- Also fixed: one dark Target reading as its whole network going unreachable ("AT&T went fully unreachable" when only one AT&T Target did); a degraded Target's hop table on an Agent Site reading "0 traces from this server" before its first trace; and the hop table's Address / PTR column breaking hostnames one character per line on a phone.
Live View
- One noisy router no longer drags Transit RTT - a transit router answering pings slowly counts less in the Transit RTT tile and its sparkline, and a real slowdown on any hop still shows.
SNMP
- Fix: a few hundred Mbps on a quiet Gateway port after its SNMP service restarted - seen after turning SNMP off and on, or a Gateway re-provision.
- Fix: a phantom multi-Gbps rate on a WireGuard, GRE, Smart Queues shaper, or Access Point
wifi0interface - the interfaces UniFi reports no link speed for. - Fix: a line-rate transfer charted a dip at its peak - a 10 Gbps transfer on a 2-second SNMP poll.
Security Audit
- Fix: an account a service keeps signed in counted against your score as an unused Local Admin (thanks @jimstrang for the report) - UniFi records only sign-ins, so an MCP server or another integration on a long-lived session reads as unused while it's in use. Console Access: Unused Local Admin Accounts no longer affects the security score, and you can Acknowledge it to dismiss it.
Alerts & Schedule
- Fix: a "Device Restarted" alert for a device that never restarted - turning SNMP off and on, or a Gateway re-provision, restarted its SNMP service and alerted "Power loss" with the reason from its last real reboot. A device Network Optimizer reaches over SSH (as reboot reasons already do) is checked now; one without SSH can still alert.
LAN Speed Test
- Fix: LAN Speed Tests failing with "iperf3 binary not found" on an On-Site Agent (native Linux and Proxmox Agent installs) - the Agent installer didn't install iperf3, so an Agent on a host without it couldn't run them. It installs iperf3 now. If you saw that error, upgrade the Agent as the callout at the top describes: the upgrade command installs iperf3 too.
- Also fixed: on a Site with two or more On-Site Agents off the Gateway, a LAN Speed Test running each direction from a different Agent.
Multi-Site
- Also fixed: the Dashboard's Sites panel drawing a degraded stretch of the 24-hour health line green.
Installation
Windows: Download the MSI installer below
Docker (Upgrade):
docker compose pull && docker compose up -dmacOS (native, recommended for accurate speed tests vs Docker Desktop):
git clone https://github.com/Ozark-Connect/NetworkOptimizer.git && cd NetworkOptimizer && ./scripts/install-macos-native.sh
# or if you already have it cloned
cd NetworkOptimizer && git pull && ./scripts/install-macos-native.shProxmox:
bash -c "$(curl -fsSL https://raw.githubusercontent.com/Ozark-Connect/NetworkOptimizer/main/scripts/proxmox/install.sh)"
# or if you just need to update
pct exec <CT_ID> -- bash -c 'cd /opt/network-optimizer && docker compose pull && docker compose up -d && docker image prune -f'For other platforms (Synology, QNAP, Unraid, native Linux) or new installations, see the Deployment Guide.
Need a hand with an install or upgrade, or something not working right? Ask on the Network Optimizer Discord. Confirmed bugs and feature requests go to GitHub Issues.