#influxdb

i'm trying to delete some data from influxdb with: DROP SERIES FROM "graphite"."netdata"."machine1"; but am getting an ERR: error parsing query: retention policy not supported at line 1, char 1

not sure if alright to ask about telegraf here, but are there any tuning options to increase speed for ingesting a backlog of log data to "catch up"? it seems to run about 1000 entries per second which is insainly slow when compared to writing even a single threaded tool

it's perfectly reasonable to read three orders of magnitude more

seems like it is throttling itself

boombatower: I think so, gimme a second to find the setting name

[agent] metric_batch_size = 1000

I set those to a million both that and other one to no avail

metric_buffer_limit

both to a million

hm

and it's not eating CPU?

using one cpu mostly and not 100%

seems like https://github.com/influxdata/telegraf/issues/3539

does the same with non custom log pattern

given I have about 1TB worth of logs I need to backfill

1000/s is completely unusable

eep

I would use it going forward

but seeing 400-800 GETs per second

so that would likely max it out

I can get roughly a million per 5 seconds in python

and million per second in php

so rather confusing go gets 1000

granted it has more abstraction, but still

well, you're not saturating the CPU

yes, but skimming code I unless there is throttling higher up I cannot see why

so there's something fundamentally wrong in there...

did you subscribe to the ticket?

yes

at first I assumed it was intentionally throttilign to keep from locking up, which may be the case

but unclear how to adjust that

no idea...

atm I'm only using telegraf+influxdb to get system metrics every 5s :P

sadly it appears I'll just have to create my own simple script

I want to add logparser

but I get 5 requests per second, so...

shouldn't be a problem :)

the config and setup is nice and makes sense

I even setup regex process to do some extra stuff, but had to go back to performance issue