Worker configuration
What you can change, what you should not, and how to change the things that are not yours.
Most of a worker's behaviour is decided by the platform, on purpose. What an operator configures is the machine's side of the arrangement.
What you set
| Setting | Notes |
|---|---|
| Name | Lowercase, digits and dashes, three to forty characters. Yours to choose; appears in the fleet register. |
| Region | Where the machine actually is. Changing it after the fact is a real change to what your measurements mean, so it goes through the platform rather than the config file. |
| Automatic updates | Off by default. Turn it on and the worker keeps itself inside the supported version window; leave it off and that is your job. |
| Resource limits | Standard Docker limits, if you want them. The worker is small enough that most operators do not bother. |
What the platform sets
The target list, the measurement interval, the batch size, and how many workers measure any given target. None of these is configurable from the worker, and that is the point: if operators could choose what and how often to measure, results from different workers would not be comparable, and consensus over incomparable measurements is arithmetic without meaning.
Changing your region
Move the machine and the region field is wrong; edit the region field and nothing about the machine has changed. Either way the platform is being told something untrue about where a measurement came from, so region changes go through the fleet register, where they are recorded with who made them and when.
If you are moving the machine, the honest sequence is to retire the worker and enrol a fresh one at the new location. The measurement history stays attached to where it was actually taken.
Version support
A worker below the supported version floor stops being counted. This is not punitive: measurement semantics change between versions, and mixing results from a version that measured differently into a consensus is how a verdict quietly becomes wrong.
With automatic updates on, this never affects you. With them off, watch the version in
vapn status.
Running more than one worker
Entirely fine, and useful when the machines are genuinely in different places or on different networks. Two workers on the same host behind the same address are not two vantage points: they share a path, so they see the same problems, and consensus treats correlated measurements as less informative than independent ones. See How consensus is calculated.
Was this page helpful?