Skip to content

@jordansissel /RESULTS.md
Created

Embed URL

HTTPS clone URL

Subversion checkout URL

You can clone with
or
.
Download ZIP
screenshot + code showing how to query logstash/elasticsearch with a graphite function.
# This requires 'pyes'
# The function goes in webapp/graphite/render/functions.py
# Don't forget to update the SeriesFunctions dict to include it.
import pyes
def logstashHits(requestContext, query):
conn = pyes.ES("semicomplete.com:9200")
start = requestContext["startTime"].isoformat()
end = requestContext["endTime"].isoformat()
# hardcode the utc offset because this vagrant box thinkts it is in CEST.
boundedquery = "@timestamp:[%s-0500 TO %s-0500] AND %s" % (start, end, query)
q = pyes.StringQuery(boundedquery).search()
q.facet.facets.append(pyes.facets.DateHistogramFacet('date_facet',
field='@timestamp',
interval='second'))
results = conn.search(query=q)
logfile = open("/tmp/x", "a")
logfile.write("q: %s\n" % boundedquery)
logfile.write("entries: %r\n" % results.facets.date_facet.entries)
logfile.write("%r\n" % requestContext["startTime"])
logfile.write("%s\n" % requestContext["startTime"])
logfile.close()
values = []
for facet in results.facets.date_facet.entries:
values.append(facet['count'])
else:
# avoid bad division?
values.append(0)
return [TimeSeries(query,
time.mktime(requestContext["startTime"].timetuple()),
time.mktime(requestContext["endTime"].timetuple()),
1, values)]
input {
twitter {
type => "twitter"
user => "USER"
password => "PASS"
keywords => [ "iphone", "samsung", "cloud" ]
}
}
output {
elasticsearch { embedded => true }
}

logstash queries graphed with graphite.

Operation: Decouple whisper from graphite.

Method: Create a graphite function that does a date histogram facet query against elasticsearch for a given query string for the time period viewed in the current graph.

Reason: graphite has some awesome math functions. Wouldn't it be cool if we could use those on logstash results?

The screenshot below is using logstash to watch the twitter stream of keywords "iphone" "apple" and "samsung" - then I graph them each, so we get an idea of popularity. As a bonus, I also do a movingAverage() on the iphone curve to show you why this is awesome.

Just to be totally clear, this implementation does not use whisper or rrd at all. The 'logstashHits()' function simply queries elasticsearch directly and produces a proper TimeSeries that graphite can use to graph! THIS IS AMAZING.

Bonus points for graphite functions being super easy to write. I used the 'sinFunction()' as a starting point since it generates its own time series.

Result:

logstash events graphed in graphite

@faja

Hi, great function:)
but if lenght of values array is not queal "start_time-end_time/interval" i get this:
https://github.com/faja/images/blob/master/logstash_graphite/1.png
instead of:
https://github.com/faja/images/blob/master/logstash_graphite/2.png
to se my fix, please take a look at fork: https://gist.github.com/faja/5031034
greetings:)

@stuart-warren

Re: your vagrant timezone issue, I found the following handy (for ubuntu at least)

config.vm.provision :shell, :inline => "echo \"Europe/London\" | sudo tee /etc/timezone && dpkg-reconfigure --frontend noninteractive tzdata"
@johntdyer

How exactly is this installed, the instructions mention something about editing the SeriesFunctions dict, but I am not sure what that means

@jtnagashima

Neat feature! Just a heads up, pyes 1.x changed the class 'StringQuery' to 'QueryStringQuery' (aparo/pyes@4497552).
So this line " q = pyes.StringQuery(boundedquery).search()" will become:
"q = pyes.QueryStringQuery(boundedquery).search()" - credit to Evan Harper.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Something went wrong with that request. Please try again.