[rrd-users] to many "data agregation" transactions at the same time

[Kristoff Bonne]

Hallo Tobias,



On 16-12-11 16:01, Tobias Oetiker wrote:
>>> the 'roll up' happens at the same time, no matter what since it is
>>> tied to the time of day. BUT the jumping from 1 block to another
>>> can differ (write position inside each rra), and since 1.4 (I
>>> think) rrdtool is randimizing this at creation time ...
>> OK, Just to be sure. Upgrading to 1.4 will NOT help my issue. Is that
>> correct?
> this depends on what your issue is, if your issue is load created
> by block transition (as the active location in the rrd file moves
> to a new block this usually causes extra data to be read). then
> files created on 1.4 will help as their write pointer is randomized
> (it will not affect existing files though ... )
> if your issue realy is the time when RRAs are updated, then there
> is no helping it ...

OK. That's a very fast and clear answer. :-)

We have the issues every 30 minutes (medium impact), 2 hours (larger
impact, server going in "waiting for IO" for several minutes) and major
impact at midnight UTC (+10 minutes "waiting for IO" resulting in
missing stats as mrtg takes to long to run).

So the answer is clear. Thanks!



Also thanks to the other people who have replied.


> cheers
> tobi
Cheerio! Kr. Bonne.