#influxdb

hi! Why is telegraf's elasticsearch input using so many measurements?

for example, a lot of elasticsearch_os_* look like they could be collapsed into multifield measurements (like in the system module) - is there any specific reason for separating them?

(I mean, I'll probably want all the elasticsearch_os_cpu_* when I'm querying for CPU usage anyway)

ktosiek: i'm not familiar with the ES plugin but i think most plugins have been rewritten to generate fewer measurements. maybe the ES plugin hasn't been updated yet

oh, so there was a change in practice?

it was awhile ago

good to know, I'm just learning influx/telegraf

maybe open a feature request to combine measurements

BTW is there a difference in storage format for multiple fields vs multiple measurements?

it might not be on anyone's radar and doesn't seems like there are any issues open about it

the underlying time series are both stored roughly the same under the hood, but the indexes holding the measurements and fields are different. i'm not super familiar on the details

ok, I'll dig into that myself, hopefuly it's in the docs :-)

thanks for help!

this has more detail but is a bit outdated https://docs.influxdata.com/influxdb/v0.12/conc...

do I understand it right that fields of a series are stored separately (kind of in columnar style)?

the storage engine docs seem to always mention "hash of series + field" as the identifier for "on-disk series"