The site probe and outside checks

A line is watched from two directions. The site probe is one device at the site that measures the line from inside. The outside checks are run by the Tenvara server over the internet. Together they tell "the site is down" apart from "the site is up but its internet is poor".

Choosing the site probe#

The probe is an ordinary device with the Tenvara agent on it. There is nothing extra to install: every up-to-date agent can be a probe and does nothing until it is chosen.

1. Open the site's Connectivity tab and find Site probe. A line at a site with no probe also offers Choose the site probe, as does the first line of that site on the Connectivity page.
2. Tenvara recommends a device. Press Use with the device's name, or Choose another device.
3. Optionally set Ping every (10 to 300 seconds), up to four internet anchors, or a gateway or ISP first hop when the automatic ones are wrong. Left empty, the values in Settings > Telephony are used.

The recommendation goes in this order: the site's network probe if it has one, then a server, then an always-on PC, and laptops last, because a laptop leaves the site and takes the probe with it. One device per site.

The probe needs outbound ping and traceroute (ICMP), DNS and HTTPS from the site. Nothing inbound.

Note: Between choosing a device and its first round, the line says Waiting for the probe's first check. That is not a fault.

What the probe measures#

Every round (20 seconds unless you change it) the probe pings three things at once:

• the LAN (gateway): the site's own router,
• the ISP (first hop): the first router past the gateway that belongs to the provider,
• the Internet: two anchors on the public internet.

Each gives latency, packet loss and jitter. The probe also looks up its public address, checks DNS and a set of web services every minute, and traces the route when the path or its latency changes.

The Site probe panel shows the hop path with each hop's latency and loss, the probe's device, state and last round, the line carrying the traffic, the public address and the interface the traffic leaves by. Below are Latency and loss charts (6 hours, 24 hours or 7 days), DNS and web services, Line changes and Traceroutes.

Telling a LAN, ISP or internet fault apart#

A round only counts as a fault when every internet anchor is lost. Then:

A gateway or ISP router that simply refuses pings while the internet works is never treated as a fault. The cause is shown on the probe panel and recorded on the outage, so you know whether to call the supplier or look at the customer's switch.

One lost ping does not make a line degraded. Quality is judged over the last few rounds: loss against the limit for the whole window (2% by default), and latency or jitter over their limits (150 ms and 30 ms) in more than half of the rounds.

When a fault is inside the site, the line shows Cannot see the line with the time the fault began. When the probe itself goes quiet, the line shows Probe silent and when it was last checked, with no stale figures left on screen.

Failover to the backup line#

The probe knows which line carries the site's traffic from its public address and its route. When the traffic moves to a backup line:

• the panel says the site is on its backup and since when,
• the backup line shows Carrying the traffic,
• Line changes records a Failover, and Back on primary when it returns.

Trace the route#

Press Trace the route on the probe panel to run a traceroute from the site, to the first internet anchor or to an address you give, such as the phone system. The hop list is kept with the site, ready to hand to the ISP. Traceroutes taken automatically are listed with the reason.

Speed tests#

Speed tests use data, so they are off until you switch them on, line by line.

1. Open the line and find Speed tests.
2. Switch on On a schedule.
3. Press Schedule and cap to set Test every (hours) and the Data cap per month in MB. Keep the cap low on metered 4G and 5G lines.

Tests run in a night window (02:00 to 05:00 in the customer's time zone, once a day by default) and only while the line is carrying the site's traffic. Test now runs one by hand. The panel shows the tests this month, the median download against the contracted speed, and the data used of the cap.

A line is marked Under-delivering this month only once it has three tests in the month: three under 70% of the contracted download, or a median under it.

Checks from outside#

As soon as a line or phone service is recorded, the server starts checking it over the internet every minute. There is nothing to set up.

• A line with a static address: a ping to each public address (up to four).
• A phone system or SIP trunk with a public address: a ping, the SIP port, the secure SIP port with the certificate it serves, and the admin console.

Checks from outside on the line's or phone service's Overview lists each check with its result, round trip, loss, the share answered over 24 hours and any certificate's expiry. From a check's menu you can Check now, Switch off or Remove, and Add a check adds a ping, TCP port or TLS port of any public address.

A few rules keep the checks honest:

• A check counts only once it has answered at least once. A router that drops pings never makes a line look down; switch that check off.
• A check is failing after three results in a row with no answer.
• While the site probe reports over a line, the probe decides the line's state. The outside checks decide it when there is no probe.
• Dynamic lines are not checked automatically. Add a check by hand if you want one.
• A phone system on a private address is watched by the site probe's SIP OPTIONS instead.

The interval, the number of failures, the ping loss that counts as degraded and how soon a certificate counts as expiring are in Settings > Telephony, under Outside checks, and most can be set per customer.