This is a Fn extension that extends the Fn API to provide some basic statistics about the operation of the Fn server.
The API uses Prometheus internally which allows statistics to be combined from multiple Fn servers, not just the one which is handling the request.
Since this is an extension it is not included in the core Fn server: to use it you need to build a custom Fn server, configured to include this extension. This API also requires a Prometheus server to be running.
There are two examples which describe how to build a custom version of the Fn server. They also describe how to start Prometheus and how to configure the custom Fn server and Prometheus to connect to one another.
The following example describes how to use the statistics API to obtain aggregated statistics from a cluster of Fn servers:
If you have Prometheus and a custom Fn server running as described in these examples, you can then deploy and run some functions and obtain information about them using the statistics API extension.
The following API call requests metric values for the past five minutes, with an interval of 30s between values.
curl 'http://localhost:8080/v1/stats'To specify a different time range and interval see Time and step parameters below
To obtain statistics for a single application hello-async-a:
curl 'http://localhost:8080/v1/apps/hello-async-a/stats'To obtain statistics for a single route hello-async-a1 in application hello-async-a:
curl 'http://localhost:8080/v1/apps/hello-async-a/routes/hello-async-a1/stats'The following API call requests metric values for the time period from starttime to endtime, with an interval of step between values. (You will need to replace the example values of starttime to endtime shown below with more recent times or you won't get any statistics.)
curl 'http://localhost:8080/v1/stats?starttime=2017-11-24T18:01:30.851Z&endtime=2017-11-24T18:11:30.849Z&step=30s'starttime and endtime should be of the form 2017-11-24T18:01:30.851Z
step should be a number followed by a time unit, such as 30s or 5m.
Here is a sample response:
{
"status":"success",
"data":{
"calls":[
{
"time":1512416119,
"value":12
},
{
"time":1512416149,
"value":20
},
{
"time":1512416179,
"value":34
},
{
"time":1512416209,
"value":41
}
],
"completed":[
{
"time":1512416119,
"value":10
},
{
"time":1512416149,
"value":15
},
{
"time":1512416179,
"value":28
},
{
"time":1512416209,
"value":35
}
],
"errors":[
{
"time":1512416119,
"value":1
},
{
"time":1512416149,
"value":3
},
{
"time":1512416179,
"value":3
},
{
"time":1512416209,
"value":3
}
],
"timeouts":[
{
"time":1512416119,
"value":1
},
{
"time":1512416149,
"value":2
},
{
"time":1512416179,
"value":3
},
{
"time":1512416209,
"value":3
}
],
"durations":[
{
"time":1512416119,
"value":13.310655658088855
},
{
"time":1512416149,
"value":14.841753632280154
},
{
"time":1512416179,
"value":34.04480194936709
},
{
"time":1512416209,
"value":45.51680103676471
}
],
"failed":[]
}
}The success element will be set to success if the API call is successful.
If the API call is unsuccessful then the success element will be set to error and an additional element error will contains a description of the failure.
The data element contains elements calls, completed, errors, timeouts and durations.
The calls element is an array of objects. Each object contains a single observation of the fn_calls counter metric at a specific time. This is a count of the number of function calls made since the server was started.
The completed element is an array of objects. Each object contains a single observation of the fn_completed counter metric at a specific time. This is a count of the number of successful function calls since the server was started.
The errors element is an array of objects. Each object contains a single observation of the fn_errors counter metric at a specific time.
This is a count of failed (but not timed out) function calls since the server was started.
If there were no errors the array may be empty.
The timeouts element is an array of objects. Each object contains a single observation of the fn_timeouts counter metric at a specific time.
This is a count of timed out function calls since the server was started.
If no function calls timed out function the array may be empty.
The durations element is an array of objects. Each object contains a single calculated value of the rolling mean fn_span_agent_submit_duration_seconds histogram metric, where the rolling mean is calculated over a period of one minute.
In addition the data element contains the element failed. This is included for backward compatibility. It is deprecated and will be removed in the future.
failed element is an array of objects. Each object contains a single observation of the fn_failed counter metric at a specific time.
This is a count of failed or timed out function calls since the server was started.
If there were no failures the array may be empty.