graphite-dev team mailing list archive

[chrismd <question175109@xxxxxxxxxxxxxxxxxxxxx>]

Question #175109 on Graphite changed:
https://answers.launchpad.net/graphite/+question/175109

    Status: Open => Answered

chrismd proposed the following answer:
Running multiple carbon instances is a great way to add capacity as long
as there isn't an underlying system resource bottleneck. That said, the
way instancing works is about to change quite dramatically. I have a
branch in progress that will be hitting trunk next week that introduces
a new configuration model for carbon. Basically instead of there being
three different types of carbon daemons with disparate features there is
going to be a single carbon-daemon.py whose behavior is completely
configurable, and there is a per-instance configuration directory. This
reduces the number of daemons you need to run in some situations and
thus reduces serialization overhead significantly.

So... if your need is immediate or you want to go with the tried and
true method, go ahead and set it up with 0.9.9. You just have to use
--instance=foo, and define a [cache:foo] section in carbon.conf for each
instance, and each has to use different ports. The next release will
include the new model so it may be good to wait a week or so until the
new stuff lands on trunk. Also, the new changes include support for the
new ceres database (webapp support is coming next) and a major cleanup
of carbon's own metric names.

I'd love to have another tester so if you go the latter route feel free
to ping me with questions/issues as much as you want :)

-- 
You received this question notification because you are a member of
graphite-dev, which is an answer contact for Graphite.