#influxdb

Hey guys, just a quick question. Am I able to load balance telegraf instances?

I'm trying to do a python client, over about 250 servers, sending a lot of metrics

I want to send all that to a load balancer VIP, and have that data round robined to telegraf instances placing the data into one influxdb cluster

why not send it directly to influxdb?

I was going to have the dev's use the statd's code they already have

and just intantiate the client with telegraf IP

figure it was easiest way to gt going

get it going*

Least amount of efforts.

If I send directly to influx then dev's need to build out a framework for writing the json out, as far as I see.

not a huge deal, but I want to make this as seamless as possible.

are you using the statsd plugin in telegraf? what is the python client doing?

right know we have statd's in python instantiated going towards a graphite

I'm under the impression I can just poing it toward telegraf

and configure telegraf to accept statsd

yup, usually telegraf would be running on each server and send batches directly to influxdb

so I wanted to actually setup a tier of telegraf servers

to handle all the metrics seperately

I typically practice seperation of concers

what if I need ot ingest more metrics then I have slaves ?

does telegraf queue and ack writes?

it will queue points and emit on the flush interval https://github.com/influxdata/telegraf/blob/mas...

awesome I'll look thru this conf file

thanks a ton man!

or maam haha

np (i'm a man)

gunnar__, discourse is more of a question/answer discussion system

now most people will say oh doesn't google groups do that too?

yeah but i don't find it too searchable and often find myself in reply mode when i'm reading

heh

does anyone do statsd with the line protocol?

thanks hoover_damm

telegraf has a statsd plugin

I am totally new to influx... can anyone give me a tip on what I am doing wrong with this derivative select on some ifInOctets counters? http://cdn.pasteraw.com/fptht571n1cwbw52wpitrx3...

why is your ocetet counter going up and then down?

good point. I feel like an idiot.. it is from an old rtg database and I think rtg does the derivative work already now that I look at it :(

so that's already a rate?

I think it probably is, yes

as far as why your derivative is null, i've never run one of those without a group by and aggregate function

so, not sure?

so i'm still sitting here fighting it. it works great if i don't try to aggregate but I am not sure what I am doing wrong with the sum/mean funcs http://cdn.pasteraw.com/9dll9whdug9n0ne30ommwcq...

logan-: is 'value' stored as integer?

I am not sure.. since the measurement schemas are not set up beforehand how do I tell what fields are typed as? I am using the influx php lib like: http://cdn.pasteraw.com/s8tnr8knqts6qk6p8jl93gg...

I guess I could cast the $row['counter'] specifically as an integer there and see how that goes, but I would also like to know how to tell what type influx thinks "value" is

i believe if you enclose the number in quation marks, it'll be treated as a string

good call.. I guess pulling it out of mysql as an integer does not necessarily mean it is typed as an integer here :) http://cdn.pasteraw.com/6adqcmao81040txfymmtw20... ... i typecasted it and apparently influx had it stored as a string before

thanks

np

php :[