Skip to content
Find file
Fetching contributors…
Cannot retrieve contributors at this time
95 lines (66 sloc) 3.65 KB

Attendance

There was about 11 people attending this first Graphite meetup

Presentation

Usage

80 : already using graphite 20 : not yet

OS

80 : Red Hat derivative 20 : Debian derivative

Dashboard Used

Venue

  • Internet : mac users had problem, it seems that it got better at some point (information needed)
  • Space : it seems to be alright for a small number, not sure if that place scales (information needed)

Discussion

  • Other backend

  • StatsD

    • Installtion : a breeze on MacOS
    • Admin : to avoid DNS resolution overhead, use a statsd instance per application (who does have 20M)
    • Admin : takes the paint from creating the datapoint in graphite first.
  • Java monitoring

    • jmx : needs to create the datapoint twice, in jmx and grapite
    • metrics.codehaus.zzz : just a jar in you load path, give you garbage colector, jetty, log4j and + for free, low overhead.
  • DevOps

    • System metrics : allows you to get your point to the other teams. An example where all the component are tracked, helped a fellow graphite ranger to pin point what was wrong and stopped the finger pointing.
  • Metrics

    • n to 1: extrapolate plenty of metrics to meaningful single data point/curve is difficult.
    • Vanity metrics : your boss/business looks at it, add value to the "graph everything proposal"
  • Package(s)

    • Graphite/dashboard : installation is a pain (especially in the RPM world)
    • Components/Depedencies : Not in the official repo (Red Hat/CentOS) or too old in the alternative repo (EPEL)
    • Sources : we end up using source (git/tarball) and compile it ourslef.
    • Repo : we all maintain our own packages repository.
    • Glue : we all use scripts that do magic, once in a while from the contrib/ folder
  • Bookmarks

  • Gdash

    • Brings a level of abstraction (reference to made to chef), nice if you don't wnat to do a 400+ characters query by hand.
    • Brings a way to pass json definition (think host definition) to a graph, help to scale.
    • Can be script to automate dashboard creation.
    • Can use some ruby inside the template ( an example inside the gdash documentation could help to promote that feature)
    • Default provided for munin and collectd view.
    • Search in the latest version, comes handy once your dropdowns go bellow the bottom of the web page.
  • Data collection

    • Collectd : seems pretty obvious candidate, still people compile it and have to rely on their version (see packaging )
    • Collectd : Default to help you started
    • Munin : script from R.I. Pienaar for front end as well as back end (munin-graphite)
    • Diamond : Mentionned, i don't know what peoples feedback was.
    • Jmx : maybe not that great (see above)
    • Statsd interface : available for a wide rage of programing languages, easy to plug to.
    • Bash / netcat : You can do a simple script to track some metrics and get the job done.
Jump to Line
Something went wrong with that request. Please try again.