the XML datastream
or the data source name in gmetad.conf?
--
Leif NixonSystems expert
National Supercomputer Centre Linkoping University
s history in Ganglia.
(We're in the middle of a major network reorganization here. You tend
to appreciate little things like this after a while...)
--
Leif NixonSystems expert
National Supercomputer Centre Linkoping University
n't then add the domainname to it.
Similar problem here; cluster nodes often don't have any concept of
a "domain".
--
Leif NixonSystems expert
National Supercomputer Centre Linkoping University
r-core-gmond, of which only the first depends on librrd.
--
Leif NixonSystems expert
National Supercomputer Centre Linkoping University
3!" | mail -s "Host alarm!" [EMAIL PROTECTED]
--
Leif NixonSystems expert
National Supercomputer Centre
piled from CVS HEAD.
I'm running a pretty stock Red Hat 8.0, with gcc 3.2 and glibc 2.2.93.
--
Leif NixonSystems expert
National Supercomputer Centre Linkoping University
olaris. 8^) Give me a day or two to clean it up, and
I'll post a patch.
I have a rough sketch for extending this to out-of-bound metrics, as
well.
--
Leif NixonSystems expert
National
Maui and inject them
into Ganglia using gmetric. Gmetad consolidates the data into rrds,
from which I generate pretty graphs, published on:
http://status.nsc.liu.se
There are some gaps in the graphs where gmonds have died on me. I
suspect the aforementioned SIGPIPE death. Let's see if t
gt; this with the world.
Great! I'll try it out next week on our 200 node cluster.
--
Leif NixonSystems expert
National Supercomputer Centre Linkoping University
ir standard paths.
I second that. /usr/local might well be shared by a set of machines,
which means RPM packages should keep their dirty little fingers off
that path.
--
Leif NixonSystems expert
-
Preston Smith <[EMAIL PROTECTED]> writes:
> On Thu, Aug 15, 2002 at 03:11:34PM +0200, Leif Nixon ([EMAIL PROTECTED])
> wrote:
> > A tentative fix would be to instead count the number of lines
> > starting with "processor" in /proc/cpuinfo.
>
> There'
c/stat and subtracts
one. Works fine for 2.4 kernels, breaks for 2.2 kernels on single-CPU
machines. In the latter case, /proc/stat only contains a single
"cpu" line.
A tentative fix would be to instead count the number of lines
starting with "processor" in /proc/cpuinfo.
Com
12 matches
Mail list logo