#influxdb

hi #influxdb, im wondering if there's any word on how clustering works in 0.9 yet

from a few weeks back but since there isn't an update I don't expect much has changed https://groups.google.com/forum/#!searchin/infl...

hi, I’ve got collectd->riemann->influxdb<-grafana, just added last two parts of that and I don’t get how to filter a query by host. ie I have (renamed) events like ‘cpu/idle’ coming from riemann into influx and i can view a graph of all of those by time - but can’t filter by host. Can anyone help me get my first useful graph built please?

matzie generally if you want to see one host, you specify it in the regular expression for the series name.

But also, you can go into "raw mode" in grafana and edit the query directly.

It just depends, is hostname part of your series name or a column in your time series?

I think I’ve just not even begun to grok the influx query lang yet. reading a bit more….

What's the deal with 0.9.0, can I start testing it out for myself yet?

i have series liek ‘cpu/idle’ and that contains a host column - i think

well if (in influx ui) I query “select * from /.*/ limit 100” I get (time, seq num,state, value, name, host) columns

ah escaping that slash is helping a lot :-)

Ok but still if I do “select * from /cpu\/idle/ limit 100” I get sensible-ish results, but select * from /cpu\/idle/ where host=“foohost” limit 100 says “cannot find column foohost” which is not what I expected

nathanielc: thanks!

ah ok, sorry for doing my learning in public - host =~ /foohost/ works. !

matzie, I could be wrong but I expect that quoted strings should be single quotes not double quotes.

In SQL'ish languages double quotes are usually reserve for quoting table and column names.

host='foohost' might be a faster way to go.

*reserved

If I'm running influxdb do I need to run statsd in front of it?

grummle, only if you need to speak the statsd protocol. influxdb natively supports a http based protocol, graphite protocol, and (I think) collectd protocol.

@shanemhansen so I can let influxdb do the aggregation thaht statsd would normally handle?

grummle, AFAIK influxdb does not do the aggregation that statsd does.

So you might want to add something like this: https://github.com/bernd/statsd-influxdb-backend to your statsd.

grummle, there's a decent writeup here: http://www.symantec.com/connect/blogs/metrics-c...

nice, thanks!

Personally all of my code speaks the graphite protocol and does it's own aggregation over a minute or so.

Then I use influxdb continuous queries to do 5m aggregates (for historical/archival stuff)

I like the idea of statsd from the standpoint of it lowers the the bar for developers, just fire off whatever you have and statsd takes care of aggregating

can you define "influxdb continuous queries "

grummle, http://influxdb.com/docs/v0.8/api/continuous_qu...

pro tip, google before you speak :)

thanks

is the current 0.9.0 estimate still March?

how do you use where to select empty columns?

Does influx have the ability to do alerts (based on predefined queries)? If not, what do people use on top of influx for alerts?

hello. i have a very basic query "select * from web.client limit 1" being called from curl, yet it just hangs there until i cancel the request. influxdb shows no sign of cpu activity and the nothing is appearing int he logs (other than the server stated up)

how can i debug this?

looks like the usual `service influxdb restart` wasnt actually restarting the process.. odd. forcly killed the process and restarted and it springs back to life again

though still no idea why it stopped in the first place