1000+ series per host causes slow performance in influxdb about telegraf HOT 4 CLOSED

@jhofeditz -- can we see your InfluxDB config file?

from telegraf.

https://gist.github.com/jhofeditz/934d8e7f296be1501707

from telegraf.

Re-reading the schema and data layout page, is the limit referring to the number of unique values per tag (host=) or the total number of unique tag series (host=,name=,serial=)? My maximum unique values per tag is only around 5,000.

If having >100,000 series is ok, then this is a InfluxDB issue and I'll open the issue there.

from telegraf.

D'oh! I might need to change the subject to "telegraf running 100x on one host causes slow performance" 😨 It turns out that on one server puppet was spawning lots of telegraf clients because /etc/init.d/telegraf status said that telegraf was stopped.

We'll watch to make sure that doesn't happen again. We wiped the database and are now going to give things another week to see how it performs. Sorry about the false alarm.

I would still like to know if the number of metrics is going to impact influxdb at scale, but this specific issue seems to be unrelated. Thanks!

from telegraf.