graphite-dev team mailing list archive

[Cody Stevens <question176255@xxxxxxxxxxxxxxxxxxxxx>]

New question #176255 on Graphite:
https://answers.launchpad.net/graphite/+question/176255

I have 2 servers now.  Each has 15k RPM drives configured in RAID 10.  I am pushing 260K metrics/minute through.  Each box has 32 G of RAM.  I noticed when looking at particularly the cache size on my graphs that periodically there is missing data.  My %util for the disks is hovering around 25% so I am not I/O bound.  The boxes may still be creating some metrics but I wouldn't think that would manifest in missing data.  I am using relays and their queues were filling up.  I have bumped the queue size to 50000 on the relay servers, figuring twice the size of my metrics/min would be good.  I am not hitting my cache size limit, I am still about 2M below it.  Any idea what could be causing this?   

Here are the values from my carbon.conf
MAX_CACHE_SIZE = 9000000 

MAX_UPDATES_PER_SECOND = 100  

MAX_CREATES_PER_MINUTE = 60 

I posted some graphs at http://www.flickr.com/photos/54244623@N07/sets/72157627980222376/

Also just another question... I noticed that my pointsPerUpdate graphs are always empty.  Is there a setting somewhere to force more pointsPerUpdate and possibly get more throughput?  

Thanks in advance for any performance voodoo you can provide.


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