Run a node

What your node can measure

A node can run three kinds of measurement. Your dashboard shows which of them are on, and this page shows how to turn on the ones that are not.

On this page

Every job a node takes is a bounded network measurement. Each kind is a capability: the node offers it only if it works on your machine, and it is only given the kinds it offers.

CapabilityWhat the node doesWhat it needs
httpMakes one web request to a public site and times each step of the answer.Nothing extra. It is on from the first start.
tcp_connectOpens a connection to a public host, times it, and closes it without sending anything.Nothing extra. It is on from the first start.
icmp_pingSends a few pings to a public host and times the replies.Permission to send ping, which the node's locked-down container does not have by itself. See Turn on ping.
dnsQueries the local ISP resolver or a fixed public resolver (1.1.1.1, 8.8.8.8, 9.9.9.9, 208.67.222.222) for a public domain.Nothing extra on Linux (hopbits-agent 0.2.5+). Your local resolver's IP is never exposed.
tracerouteProbes successive router hops toward a public host over unprivileged UDP (IP_RECVERR) or ICMP.Nothing extra on Linux (hopbits-agent 0.2.5+). Private and Hop-1 residential routers are redacted.
mtrRuns 3 to 10 probe cycles across each hop to measure path packet loss and jitter.Nothing extra on Linux (hopbits-agent 0.2.5+). Private and Hop-1 residential routers are redacted.

With all three on, your node can be given every kind of job the network has.

Where to see them

In the operator dashboard, each node under Connected nodes has three marks in the Capabilities column: HTTP, TCP and Ping. A filled mark is on and an empty one is off. An empty mark is a link that brings you to this page.

The node tests itself every time it connects and again every six hours, and reports what works. The dashboard shows the latest report while the node is online. For a node that is offline it shows no marks as on, because nothing is known.

Turn on ping

HTTP and TCP work straight away. Ping is the one that can show as off. The node's container runs with every Linux capability dropped, so it can only send ping when the system lets ordinary programs do so. Many Linux systems do by default, and some do not.

This is the default, and what the installer sets up. Allow ping on the machine that runs the node:

Linux shell
sudo sysctl -w net.ipv4.ping_group_range="0 2147483647"

To keep it after a reboot, save the same setting in a file:

Linux shell
echo 'net.ipv4.ping_group_range = 0 2147483647' | sudo tee /etc/sysctl.d/99-hopbits-ping.conf

Then restart the node, so that it tests itself again at once:

Linux shell
sudo docker restart hopbits-agent

Once the node has reconnected, the Ping mark fills in on the dashboard. Without a restart, the node notices by itself at its next self-test, which can be up to six hours away.

When HTTP or TCP is off

These two need nothing from the machine. If one of them is off, the usual reason is that it was switched off on purpose with HOPBITS_DISABLE_KINDS. Start the node again without that setting, or without that kind in its list. Configuration lists the settings.

Switch a capability off

What your node runs is your choice. To leave one kind of measurement to other nodes, name it in HOPBITS_DISABLE_KINDS, for example HOPBITS_DISABLE_KINDS=icmp_ping. At least one kind has to stay on.

IPv4 and IPv6

Every capability works over IPv4. A node also takes jobs for IPv6 targets when its machine has a public IPv6 address.