Hi,
I used to work at Flickr where we use(d) a standard Ganglia install to
great effect.
I've since left Flickr and landed at Stamen (Design) where we've been
using a custom RRD store with RRAs that look like this:
RRAs=\
» RRA:MIN:0.5:1:480·\
» RRA:MIN:0.5:5:2016·\
» RRA:MIN:0.5:60:720·\
» RRA:MIN:0.5:1440:365·\
» RRA:AVERAGE:0.5:1:480·\
» RRA:AVERAGE:0.5:5:2016·\
» RRA:AVERAGE:0.5:60:720·\
» RRA:AVERAGE:0.5:1440:365·\
» RRA:MAX:0.5:1:480·\
» RRA:MAX:0.5:5:2016·\
» RRA:MAX:0.5:60:720·\
» RRA:MAX:0.5:1440:365
So, basically we collect stuff at 60 second interval and generates
aggregates in 1 minute, 5 minute, 12 hour and daily chunks.
Recently, I've been setting up Ganglia for a number of projects and out
of the box it works great.
We'd also like to keep generating the same RRAs that we have, alongside
the Ganglia, defaults. We have a bunch of custom graphs that we'd like
to keep using (but route all the collection and aggregation through
Ganglia).
This is where I seem to be running in to problems. There are two
specifically:
1) None of the (custom) RRAs seems to collect more than an hour's (maybe
two) worth of data. By which I mean all the graphs look like this even
when gmetad's been running for a day or two:
http://aaronland.info/rrd.png
They just never get any longer which suggests that my RRA definitions
are incorrect but I'm not sure why, exactly.
2) There seems to be some hard limit on the number of RRAs I can define?
Maybe?
Here are the details:
These are the RRAs I'm using, assuming that since the default (rrdtool)
step for Ganglia is 15 (seconds) I should just be able to multiply
everything by 4.
It's not entirely clear to me whether I need to do the same for the
rows. I don't think so but I am a bit dense that way, sometimes...
"RRA:MIN:0.5:4:480" "RRA:MAX:0.5:4:480" "RRA:AVERAGE:0.5:4:480" \
"RRA:MIN:0.5:20:2016" "RRA:MAX:0.5:20:2016" "RRA:AVERAGE:0.5:20:2016" \
"RRA:MIN:0.5:240:720" "RRA:MAX:0.5:240:720" "RRA:AVERAGE:0.5:240:720" \
"RRA:MIN:0.5:5760:365" "RRA:MAX:0.5:5760:1460" "RRA:AVERAGE:0.5:5760:365" \
"RRA:AVERAGE:0.5:1:244" "RRA:AVERAGE:0.5:24:244" "RRA:AVERAGE:0.5:168:244"
Then there's the second problem where, given the above I can never seem
to get the last two default RRAs to load properly. Specifically:
# "RRA:AVERAGE:0.5:672:244"
# "RRA:AVERAGE:0.5:5760:374"
They always fail with the following error:
RRD_create: can't parse argument '"RRA:AVERAGE:0.5:672:244"
...Which seems to be triggered by the corresponding "else" block to this
in rrd_create.c (~ ln. 166):
if (strncmp(argv[i], "DS:", 3) == 0) {
I'm not really a C programmer so at that point I decided I was about to
fall in to a twisty maze of stuff I would gradually understand less and
less.
Any thoughts or (gentle :-) cluebats on all of this would be most welcome.
Thanks,
------------------------------------------------------------------------------
_______________________________________________
Ganglia-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ganglia-general
