Installation, configuration, and operations · Applies to SLA Probe 1.1 (adds traceroute support)
About the SLA Probe
The NetOp SLA Probe is a lightweight software agent that runs on-premises on a commercial off-the-shelf (COTS) server or virtual machine. It continuously measures network Service Level Agreement (SLA) metrics against a configurable list of targets and exports the results to a central NetOp Agent server for aggregation and analysis.
Targets can be public services (Zoom, Office 365), DNS resolvers, internal servers, or any IP address or hostname of interest. For each target, the Probe measures:
- Latency (ms)
- Jitter (ms) — variation in latency between consecutive checks
- Packet loss (%)
- MOS (Mean Opinion Score, 1–4.5) — a single composite score approximating voice/call quality, derived from latency, jitter, and loss
Each Probe runs independently, storing results locally and exporting them to a central NetOp Agent server, so data from many probes across many sites can be aggregated in one place.
Audience: technical personnel involved in the design, specification, or deployment of the NetOp SLA Probe at a customer site.
Related article: NetOp Agent Installation Manual
Supported check types (protocols)
| Protocol | What it does |
|---|---|
icmp |
Standard ping |
http |
HTTP/HTTPS GET request; measures response time and status code, broken down into DNS lookup, TCP connect, TLS handshake, and time-to-first-byte (see SLA profile targets for the fields that control this) |
tcp |
TCP connect test to a specific host and port |
dns |
DNS resolution timing against a specified nameserver |
Probe VM requirements
Network requirements
- Management network — a reachable network interface to the NetOp Agent server
- Outbound network — access to whatever SLA targets are configured
- The outbound network connection can be wired or Wi-Fi
- The outbound and management network interfaces can be the same physical interface
Minimum system requirements
| Requirement | Minimum specification |
|---|---|
| Operating system | Linux server, Ubuntu 24.04 LTS |
| Processor | 1 GHz or faster, 64-bit, 1 vCore |
| Memory | 2 GB RAM |
| Storage | 30 GB disk space |
Installation (native Linux)
-
From the Linux home folder, download the installation package:
wget https://storage.googleapis.com/netop-probe/release/netop-probe.run
-
Change the file permissions:
chmod +x netop-probe.run
-
Run the Probe installation script:
./netop-probe.run
- Follow the instructions of the installation script. When installation finishes, you should see the message "installation complete!"
Configuration (netop-sla.yaml)
The configuration file is located at ~/netop-probe/config/netop-sla.yaml. It is organized into sections, and every section should be reviewed and set for each installation.
Note: After any change to the configuration file, restart the Probe for the new settings to take effect (see Operational commands).
Global
global: default_interval: 30
default_interval — fallback check interval, in seconds, used for any target that doesn't specify its own interval.
Traceroute — schedule and method
These settings apply to every target's traceroute companion check, which runs automatically alongside the regular SLA check unless a target opts out (see SLA profile targets).
global: traceroute_interval: 5 traceroute_method: "tcp" traceroute_tcp_port: 443 traceroute_max_hops: 30 traceroute_queries: 3 traceroute_timeout: 2
| Setting | What it does (default) |
|---|---|
traceroute_interval |
Minutes between traceroute runs for every target (minutes, not seconds, unlike default_interval). Default: 5. |
traceroute_method |
Probe method traceroute uses: "icmp" | "udp" | "tcp". Default: "tcp". |
traceroute_tcp_port |
Fallback TCP port for tcp-mode traceroute when a target has no port of its own. Default: 443. |
traceroute_max_hops |
Give up after this many hops. Default: 30. |
traceroute_queries |
Probes sent per hop on Linux/macOS (Windows tracert always sends 3). Default: 3. |
traceroute_timeout |
Seconds to wait per probe. Default: 2. |
Traceroute — adaptive method rotation
When traceroute_adaptive_method is enabled (the default), the Probe doesn't stick to a single fixed method for every target. It tries icmp, udp, and tcp against each target, scores how reliably each one reaches the destination, and automatically prefers whichever method works best for that target's network path. This is useful because some networks or firewalls block or rate-limit one method (ICMP is a common example) while allowing another. The learned preferences and scores are kept locally per target and are not exported.
global: traceroute_adaptive_method: true traceroute_adaptive_bad_quality: 0.6 traceroute_adaptive_min_samples: 5 traceroute_adaptive_alpha: 0.2 traceroute_adaptive_switch_margin: 0.15 traceroute_method_reset: false
| Setting | What it does (default) |
|---|---|
traceroute_adaptive_method |
true = automatically rotate icmp/udp/tcp per target based on which works best; false = always use traceroute_method. Default: true. |
traceroute_adaptive_bad_quality |
Quality score (0–1) below which a method is considered underperforming and an alternative is tried. Default: 0.6. |
traceroute_adaptive_min_samples |
Samples collected for a method before its learned score is trusted enough to act on. Default: 5. |
traceroute_adaptive_alpha |
EWMA smoothing factor used to update each method's learned score after every run. Default: 0.2. |
traceroute_adaptive_switch_margin |
How much better an alternative method's score must be before the Probe switches its preference to it, to avoid flapping between methods. Default: 0.15. |
traceroute_method_reset |
One-time command, not a steady-state setting: set to true, restart the Probe once to clear all learned method preferences, then set back to false. Default: false. |
Traceroute — culprit detection
On every traceroute run, the Probe flags a "culprit" hop — the point in the path most likely responsible for latency — using three independent checks:
- a sudden latency jump from one hop to the next
- a hop running well above its own learned historical baseline
- a sanity check on hop 1 (your gateway), which should normally answer almost instantly
global: traceroute_culprit_threshold_ms: 20 traceroute_baseline_threshold_ms: 100 traceroute_baseline_min_samples: 5 traceroute_hop1_sanity_ms: 10 traceroute_baseline_reset: false
| Setting | What it does (default) |
|---|---|
traceroute_culprit_threshold_ms |
Hop-to-hop latency jump, in ms, that flags a hop as the culprit. Default: 20. |
traceroute_baseline_threshold_ms |
Latency above a hop's own learned baseline, in ms, that flags it as the culprit. Default: 100. |
traceroute_baseline_min_samples |
Samples collected before a hop's baseline is trusted enough to use. Default: 5. |
traceroute_hop1_sanity_ms |
Hop 1 (your gateway) latency above this, in ms, is always flagged regardless of baseline. Default: 10. |
traceroute_baseline_reset |
One-time command, not a steady-state setting: set to true, restart the Probe once to clear all learned hop baselines, then set back to false. Default: false. |
Host
host: interface_name: eth0
interface_name — the local outbound network interface that SLA test traffic routes through (for example, eth0 on Linux, en0 on macOS).
To find which interface the SLA outbound traffic is routed through, use:
ifconfig -a ip link ip route
NetOp Agent server
redis: server: 192.168.1.132
server — address of the central NetOp Agent server this Probe exports its data to.
SLA profile targets
A list of checks to run. Each entry requires, at minimum, name, protocol, and target; other fields depend on the protocol.
| Field | Applies to | Notes |
|---|---|---|
name |
all | Unique label for this check |
protocol |
all |
icmp, http, tcp, or dns
|
target |
all | IP address, hostname, or URL, depending on protocol |
interval |
all | Seconds between checks (overrides global.default_interval) |
count |
icmp | Number of ping packets per check |
timeout |
all | Per-check timeout, in seconds |
port |
tcp | Port to connect to |
verify_ssl |
http | Whether to validate TLS certificates |
dns_server |
dns | Nameserver to query against |
traceroute |
all | Set to false to disable the traceroute companion check for this specific target. Default: true (runs for every target). |
http_detailed_timing |
http |
true (default) uses the phase-timed check, reporting dns_ms / tcp_connect_ms / tls_ms / ttfb_ms. Set to false to use the original single-number check instead — for example, for an HTTP/2-only endpoint that reacts oddly to the hand-rolled client. |
follow_redirects |
http | Whether to follow HTTP redirects. Default: false. Setting this to true switches the target to the single-number check, since the detailed-timing path doesn't follow redirects. |
user_agent |
http | Custom User-Agent header string to send. Defaults to a standard desktop Chrome user agent string. |
Any global.traceroute_* setting from the Global section can also be set directly on an individual target to override the global value for that check only.
To add new SLA targets, edit ~/netop-probe/config/netop-sla.yaml, then restart the Probe.
Probe operational commands
Use ./probectl.sh with one of the following commands:
| Command | What it does |
|---|---|
./probectl.sh start |
Start the Probe |
./probectl.sh stop |
Stop the Probe |
./probectl.sh restart |
Restart the Probe |
./probectl.sh status |
Show whether the Probe is running |
./probectl.sh logs |
Follow the Probe's logs (Ctrl+C to stop watching) |
./probectl.sh uninstall |
Stop the Probe and remove its containers, image, and data |
What to do after you make a change
- Restart the Probe software after any configuration change (
./probectl.sh restart). - Wait about an hour, then run Sync from NetOp → Operations.
- It takes a few minutes for the changes to appear.
Have feedback on this article? Email info@netop.cloud. Learn more at netop.ai.
Comments
0 comments
Please sign in to leave a comment.