At my work, Say Media, we have a group shepherding tasks and plans in the 'devops' style. Franck and myself are this group, with input and assistance from others.
As a starting point we’re assuming that tools must already exist, and that we should only rebuild them from scratch if the barrier to entry is too high. Reuse, contribute, and participate in the Open Source landscape. That said, we will build tools to glue things together. My intention in this and the coming posts is to share our research, decisions, and rationale for tools we are looking at.
We started off this year by rolling out a replacement for system and application metric collection. We didn't perform any sort of experiment to test systems and see what worked the best. Instead we dove right in and started replacing Ganglia with a more discrete system comprised of Collectd, Carbon, Whisper, and Graphite (the latter three which are often considered a single item.) Statsd got added to collect counters, and much of this setup has been tweaked, tuned, and twisted to do what we expect. Finally we looked at riemann for analysis and triggering alerts.
Starting now I’ll be writing posts to document our Collectd, Graphite, Statsd, and Riemann experience. There will certainly be more, but stay tuned and ask me to share more if you want.
Comments