graphite-dev team mailing list archive

[chrismd <question170527@xxxxxxxxxxxxxxxxxxxxx>]

Question #170527 on Graphite changed:
https://answers.launchpad.net/graphite/+question/170527

chrismd proposed the following answer:
Dyna, the listener log is showing you how frequently clients are
connecting to carbon. So its processing 1.5 connections per second ,
which is just how fast your clients must be connecting to it. Look at
the updates.log to see how fast it is writing datapoints to disk.

The best thing to do is to figure out where the bottleneck is
specifically. You mention the CPU usage spikes to 100% when you send a
batch of data, that is somewhat expected but it should be very short-
lived, if CPU usage is consistently high there may be some other problem
going on. If CPU usage is not consistently high I would expect the
bottleneck to be I/O utilization, what does iostat show? If the disk is
more than 50% utilized I would advise setting MAX_UPDATES_PER_SECOND to
something lower like 500, the default of 1000 is too high for most
systems and is going to be lower in the next release. It turns out that
trying to write too quickly can cause worse performance than relying
more on buffering.

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