graphite-dev team mailing list archive

[chrismd <question144005@xxxxxxxxxxxxxxxxxxxxx>]

Question #144005 on Graphite changed:
https://answers.launchpad.net/graphite/+question/144005

    Status: Open => Needs information

chrismd requested for more information:
Take a look at carbon's performance metrics, under carbon.agents.*

The key metrics are:

avgUpdateTime - this is the average write() call latency, should be consistently very low.
cache.size - this is how many datapoints are in carbon's cache. This will increase as avgUpdateTime increases.
committedPoints - this is how many datapoints are being written each minute. It will drop if avgUpdateTime increases.
creates - this is how many new wsp files are being created each minute, it is generally capped at some configured limit. If you have a ton of new metrics coming in but the creation rate is too low the new metrics will cause the cache to grow excessively.
cpuUsage - this increases with cache.queries and cache.size

The larger the cache size is, the more cpu use carbon will eat up
because cache queries will have larger results (which cost cpu to
serialize).

A few relevant settings in your carbon.conf to check:

MAX_CACHE_SIZE
MAX_UPDATES_PER_SECOND
MAX_CREATES_PER_MINUTE

Let me know where these config/metric values are and I can give you some
more insight as to what may be causing the problem. Also, check and see
how much available memory the system has and how much of it is being
used by the carbon-cache process.

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