I'm a bit confused by this bit about histogram data:
|
Send [**Wavefront data format**](wavefront_data_format.html) histogram data only to a minute, hour, or day port. |
|
* If you send Wavefront data format histogram data to the distribution port, the points are rejected as invalid input format and logged. |
|
* If you send Wavefront data format histogram data to port 2878 (instead of a min, hour, or day port), the data is not ingested as histogram data but as regular Wavefront data format metrics. |
It makes the most sense if I interpret "Wavefront data format histogram data" to mean multiple samples in the "metrics" data format, i.e. this:
my.metric 10 <t1> <source>
my.metric 20 <t2> <source>
my.metric 20 <t3> <source>
And if I interpret "histogram distribution data format histogram data" to mean data like this:
{!M | !H | !D} [<timestamp>] #<points> <metricValue> [... #<points> <metricValue>]
<metricName> source=<source> [<pointTagKey1>=<value1> ... <pointTagKeyn>=<valuen>]
With those two assumptions, I think I grok the lines linked above.
I am hesitant about those assumptions because the explanation of "Wavefront data format histogram data" is a link to the page that explains the syntax of both the regular metrics format and the histogram format. The term "Wavefront data format" in the table above those lines also seems intended to identify a specific data format, but again, it links to the same page which describes the syntax for multiple data formats.
Are my assumptions correct? And if so, could I suggest altering the terminology to reduce confusion? I think we should use shorter, more distinct names to distinguish these two concepts. The first type above could be referred to as the "metrics format", correct? That's what it is, albeit expecting many samples within short time ranges. The second type mentioned above could be referred to simply as the "histogram format". In either case, "Wavefront data format" is far too vague and could reasonably be interpreted to mean the superset of all data formats supported by Wavefront, rather than one the format of one specific type of data.