graphite-dev team mailing list archive

[Dyna <question170527@xxxxxxxxxxxxxxxxxxxxx>]

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

Dyna gave more information on the question:
Ah forgot a few things:

>Are you able to rate limit the sending process or generally what level of
>control do you have over this process?

I've total control of that one, later on when we swith to feed live data
I assume it will be less of control.  But this is just a test, it's easy
to import data.

>If your ability to modify this sending process is limited, do you have
>access to extract and format the data outside of the sending process?

Yes I can limit it, just put a wait stage and I can say send 10 data
points evey minute, but that would really cut the mustard.

>What are your acceptable levels of utilization of memory/CPU/IO?
Well I don't expec the process to eat all my mem. If so I don't fancy creating swap mem since the only thing that will happen is that the box will start to swap and come to a crawl.  A cach is a-ok but you really need to be able to set it level say 50% of the RAM.  It's should be an abolute priority to flush write data to disk as fast as possible, when you start getting close to your limit. Submitting eight datapoints a second from one server should not use 99% of the CPU, rather just a few%.  When you start feeding data from 100 servers, then you I could see util of mem and cpu being very high. 

Aprox how many total data points per metric do you wish to import?
It's as said eight datapoints every five minutes for five years.. 

Cheers Dyna

PS: Sorry for my typos it's not easy to type with a big bandage..

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