mark-
i've seen this behavior on the machine running the ganglia demo page but
it's just a p2 with 128 mbs of memory (soon to be upgraded).
i'm rewriting gmetad in C right now and will be incorporating it into the
monitoring-core distribution soon. the biggest bottleneck right now with
gmetad is disk I/O. keep in mind the load on linux is a measure of the
number of running processes but also those processes in I/O wait. gmetad
is writing to about 25 files per host every 15 seconds or so. the next
generation of gmetad will not be nearly as i/o intensive.
as a trick to make gmetad work better for you.. create a ramdisk to write
the round-robin databases to.
here are the steps (i'm assuming you installed gmetad in the default
location)
1. find out how much space your round-robin databases are taking right now
by doing the following
a. # cd /usr/local/gmetad/rrds
b. # du -sk .
80384 .
it's important to note the side of the round-robin databases remain
constant over time and don't grow in size. of course, if you increase
the number of databases (by monitoring more hosts or metrics) then this
number will increase. in this example (taken from the ganglia demo
machine), we are monitoring 117 hosts for a total of over 3000 rrds
files with only 78 mbs of disk space.
2. create a ramdisk image file at least as big as the space you need (i'd
double it... 80384*2= 160768)
dd if=/dev/zero of=/root/rrd-ramdisk.img bs=1k count=160768
3. mke2fs -F -vm0 /root/rrd-ramdisk.img
4. /etc/rc.d/init.d/gmetad stop
5. mv /usr/local/gmetad/rrds /usr/local/gmetad/rrds.orig
6. mkdir /usr/local/gmetad/rrds
7. mount -o loop /root/rrd-ramdisk.img /usr/local/gmetad/rrds
8. copy your round-robin databases to the new RAM disk...
(cd /usr/local/gmetad/rrds.orig; tar -cf - .) | \
(cd /usr/local/gmetad/rrds; tar -xvf -)
9. /etc/rc.d/init.d/gmetad start
if you want to see a site which uses this RAM disk trick (and invented
this trick too) take a look at http://meta.rocksclusters.org/. they are
monitoring over 450 hosts using this method quite comfortably.
one important note... since the data is being written to RAM and not the
disk.. it of course will be lost on reboot. if you want to keep the
round-robin databases long-term.. you will need to setup a cron job which
saves the data from the RAM disk to the physical disk and then writes it
back on reboot.
i hope this helps. i'm going to focus much more attention on gmetad in
the next few days and i'm sure you'll find the C version of gmetad much
much more easier to install and much more efficient to run.
good luck!
-matt
Yesterday, markp wrote forth saying...
> Is anyone experiencing a high load with gmetad? I've run this daemon on
> a high end intel 933mhz dual proc machine with 1gb of memory and RH
> 7.2. Loads get and stay as high as 3. I get worse results on single
> processor machines, loads as high as 6.7 Kill the daemon and it drops
> back to normal. Is it supposed to be such a resource hog? I ran the
> old web-frontend and didn't have any problems.
>
>
>
>
> -------------------------------------------------------
> This sf.net email is sponsored by: OSDN - Tired of that same old
> cell phone? Get a new here for FREE!
> https://www.inphonic.com/r.asp?r=sourceforge1&refcode1=vs3390
> _______________________________________________
> Ganglia-general mailing list
> Ganglia-general@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/ganglia-general
>
