graphite-dev team mailing list archive

[Drew Trieger <question227742@xxxxxxxxxxxxxxxxxxxxx>]

New question #227742 on Graphite:
https://answers.launchpad.net/graphite/+question/227742

I have 0.9.9 running on a single vm which can manage at peak about 1m metrics/minute from a single source blasted in to port 2003 doing no aggregation.  Steady state flow is 200k/min.  This setup is about a year old.  The cache is set to be small, 1000000, and at peak python cpu is just over 100%, at steady state is around 45% of a single cpu. Python 2.6

I installed (thanks Leinartas) latest and greatest on a new vm (0.9.x  using git, vm running centos 6.2) and rsynced a consistent copy of /opt/graphite/storage there.  
It worked, showing the old metrics, fine.
I then redirected my live metrics flow to it and immediately (not even after filling the cache) the carbon processes (python) cpu goes to 105% and since flow control is on, the consummation rate is quite low, 30k/min or so.  It never speeds up.

Its not IO bound, either in reading the whisper files nor in writing updates.  

My guess is that there is some excessive regex or processing going on, but I see nothing interesting in the logs.

Anyone have any ideas what to do to debug?
(try another version?  Are there any cpu-heavy features in carbon now?)

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