External collectors are executables or scripts that scollector invokes, collects output from, and forwards data points on to the server like any other metric. These can be binaries, .sh/.py/.rb/.cmd/.bat/.ps1 scripts, or even .jar files ( See InitPrograms function for implementation details ). External collectors are a great way to get started collecting data, but when possible it is recommended for applications to send data directly to Bosun or to update scollector so that it natively supports additional systems.
The ColDir configuration key specifies the external collector directory, which is usually set to something like /opt/scollector/collectors in Linux or C:\Program Files\scollector\collectors in Windows. It should contain numbered directories just like the ones used in OpenTSDB tcollector. Each directory represents how often scollector will try to invoke the collectors in that folder (example: 60 = every 60 seconds). Use a directory named 0 for any executables or scripts that will run continuously and create output on their own schedule. Any non-numeric named directories will be ignored, and a lib and etc directory are often used for library and config data shared by all collectors.
The simple output format can be use to send raw metrics to scollector using the standard output stream. You won’t be able to specify units, counter vs gauge, or descriptions, but those can always be specified manually in graphs/expressions or added later if desired.
Metric is the name of the metric you want to create.
Timestamp is in Unix format (seconds since epoch).
Value is the floating point value that will be used for that data point.
Tags are optional. A host tag is automatically added, but overridden if specified.
Errors anything written to the standard error stream is included in the scollector logs (syslog on Linux, event log on Windows).
If you want to include metadata with your external collector metrics you can send JSON serialized instances of the opentsdb.DataPoint struct for metric data and the metadata.Metasend struct for metadata. The JSON data would be streamed over standard output and look similar to the output listed below. It starts with metadata (only needed once, as scollector will re-send as needed) and then continues sending metrics. The 4th line is a log message which was sent to the standard error stream.