Summary

I’m developing a custom python.d plugin that collects integer counter values every second

However, when I query the data via Netdata’s API (/api/v2/data or /api/v3/data),
I sometimes see fractional values such as:

[1760926238,
    [689,0,0,1]],
[1760926237,
    [687,0,0,1]],
[1760926236,
    [686.0026,0,0,1]],
[1760926235,
    [686,0,0,1]],
[1760926234,
    [685,0,0,1]],
[1760926233,
    [684,0,0,1]],
[1760926232,
    [683,0,0,1]],
[1760926231,
    [681,0,0,1]],
[1760926230,
    [680.002,0,0,1]],
[1760926229,
    [680,0,0,1]],
[1760926228,
    [679,0,0,1]],

from bases.FrameworkServices.SimpleService import SimpleService

update_every = 1
priority = 2

ORDER = [
    'test'
]

CHARTS = {
    'test': {
        'options': [None, 'Bandwidth', 'bytes/s', 'traffic', 'test.test', 'line'],
        'lines': [
            ['test', 'test', 'absolute']
        ]
    }
}

class Service(SimpleService):
    def __init__(self, configuration=None, name=None):
        SimpleService.__init__(self, configuration=configuration, name=name)
        self.counter = 0
        
    def check(self):
        return True

    def _get_data(self):
        self.counter += 1
        data = {
            'test': int(self.counter)
        }
        return data

Expected Behavior

For absolute dimensions, I would like the API to return exact raw integer samples
as stored in the DB (tier 0), without any fractional interpolation or partial-bucket adjustment.

Actual Behavior

Occasionally small fractional offsets appear in the API results,
even when querying with:

http://127.0.0.1:19999/api/v3/data?contexts=test&options=raw,seconds,natural-points,unaligned,selected-tier&tier=0&time_group=sum
options=raw,seconds,natural-points,unaligned,selected-tier
tier=0
time_group=sum