Worker health

How to tell whether your worker is doing anything useful, and what the numbers mean.

A worker can be running, enrolled and completely useless. These are the things worth checking, and the ones that look alarming and are not.

The four signals

A growing queue with current heartbeats means the worker can reach the platform to say hello and not to upload, which usually points at request size or an intermittent path rather than a firewall.

Things that look wrong and are not

Most targets unreachable. Plenty of hosts never answer pings. What matters is whether your worker sees silence that other workers do not.

Your worker disagreeing with the consensus. Sometimes it is right and everybody else is far away from a problem it can see. Persistent disagreement in situations where nobody else sees anything is the case worth looking at.

No visible influence after enrolling. Expected. Influence is earned by having a record.

Brief silences. Reboots and network blips. A worker returns to active by itself when heartbeats resume; nothing needs doing.

Things that are wrong

Enrolled, heartbeating, never uploaded. Nearly always ICMP. See Worker connectivity.

Silent for days. The machine is off, the container has stopped, or the network is gone. vapn status and vapn logs on the host.

Suspended. Staff have stopped it contributing, and you will have been told why. Its history is kept and suspension is reversible.

The three commands

vapn status     # enrolment, snapshot age, last upload, version
vapn logs       # what it has actually been doing
vapn doctor     # the install checks, re-run against the machine as it is now

vapn doctor first when something changed. It catches the case where something true at install time has stopped being true, which is the majority of "it was working yesterday".