at this point the 2GHz 8 core amd box, with 8 drives and 32GB memory has a
load that hangs out around .7 most of the time and spikes up to 4 for
maybe 5 minutes when the pdflush kicks in every 2 hours. cpu around 12%
system and 12% user.
i tried the tmpfs thing (actually /dev/ramN), but decided syncing back and
forth to disks to preserve data across reboots was suboptimal. never
tried rrdcached.
i'm pretty sure i could double if i were not keeping MAX archives along
with AVERAGE... at this point i have about 19GB of rrds.
-scott
On Thu, 18 Feb 2010, Bernard Li wrote:
Hi Scott:
On Wed, Feb 17, 2010 at 11:47 AM, Scott Dworkis <[email protected]> wrote:
a clue that you are bumping into io cache is also a lot of dropouts in
metrics graphs, a lot of cpu io-wait time, and a consistent showing of
pdflush ranked high in bin/top. i suspect there are some risks with this
cache tuning, so use with caution.
Most users run into this I/O related issue at around 30k metrics. I
am curious what the resource usage of your gmetad is at 150k metrics
after your tweaks.
User usually work around the issue by:
1) Putting the rrd files in tmpfs
2) Using rrdcached
Would be interested to see where the next scalability wall is.
Cheers,
Bernard
------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
Ganglia-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ganglia-general
