pdjr-skplugin-interfacewatchdog v1.0.0
pdjr-skplugin-interfacewatchdog
Signal K interface activity watchdog.
Description
interfacewatchdog monitors the activity of a specified Signal K interface waiting for the connection rate to fall below some specified threshold. If this happens, the plugin writes a message to the server log, issues a notification and optionally restarts the host Signal K server.
The plugin was designed to monitor interfaces associated with a data connection, but can be used against any interface listed in the server dashboard connection panel.
Configuration
The plugin recognises the following configuration properties.
| Property | Description | Default value |
|---|---|---|
| interface | The Signal K interface that should be monitored. | 'n2k-on-ve.can-socket' |
| threshold | The data rate (in deltas per second) at or below which the plugin should act. | 0 |
| restart | Whether or not the plugin should restart the Signal K host when throughput drops below the specified 'Threshold' value. | true |
| notificationpath | The path under vessels.self on which the plugin should issue alarm notifications. | 'notifications.interfacewatchdog' |
Operation
The plugin remains idle until throughput appears on interface. This prevents the plugin causing repeated, immediate, restarts on a server on which interface is disconnected or otherwise dead.
Once activity is detected on interface the plugin checks the throughput statistic reported by Signal K each time the server issues a 'serverevent' of type 'SERVERSTATISTICS' (typically every four or five seconds). You can view these activity values in the Signal K dashboard.
If the detected throughput is less than or equal to threshold then an alarm notification is issued on notificationpath and, if restart is true, the host Node process is killed.
Note that:
If restart is not true then any previously issued alarm notification is cleared if throughput again rises above threshold.
If and when the Signal K Node process is killed it will only restart automatically if the host operating system's process manager is configured for this behaviour.
The kill signal is issued approximately one second after the alarm notification is issue on notificationpath: this delay is designed to allow an alarm handler or annunciator to detect the alarm condition and do its thing.
The event handler which detects interface throughput cannot update plugin status information in the Signal K Dashboard, so the only plugin status messages you will see on the server dashboard are those associated with plugin initialisation.
Background
The plugin was written as a tool to help diagnose a problem on my own vessel where a suspected buggy N2K device is occasionally issuing a broken PGN which in-turn causes Signal K's CAN interface driver (in my case 'canboatjs') to lock-up.
Automatically restarting Signal K when an interface lock-up is detected stops the problem becoming a major issue until the inferred 'canboatjs' bug can be diagnosed and fixed.
Author
Paul Reeve <preeve_at_pdjr_dot_eu>
2 years ago