graphite-dev team mailing list archive

[Drew Trieger <question186672@xxxxxxxxxxxxxxxxxxxxx>]

Question #186672 on Graphite changed:
https://answers.launchpad.net/graphite/+question/186672

Drew Trieger posted a new comment:
I should have added what the impact is.
It means that when I detach graphite from prod for maintenance, capacity change, etc. a backlog of metrics build up in my queuing mechanism, good.
After an hour, I bring graphite back into the flow (now with more spindles or more ram or ...),
it takes the flood of queued up metrics, processes them correctly, mostly spewing into cache.
Cache hits the limit and the inflow stops.
The draining of the cache slows down dramatically as it drains, as the # of points per update shrinks.  
As a result, it might take an hour before it gets actually zero, getting less and less efficient as it goes.
Once it gets to zero, it takes metrics again, but its an hours worth by this time and so the process repeats, and it can never catch up.

So this limits the amount of time I can have graphite decoupled for
maintenance.

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