[rrd-users] Re: Runaway rrdtool process
anders_nygren2002 at yahoo.se
Sun Feb 8 23:03:01 MET 2004
On Sunday 08 February 2004 09:11, you wrote:
> On Fri, Feb 06, 2004 at 11:46:31AM -0600, Anders Nygren wrote:
> > When testing I have 1 rrdtool process, and send 15 updates per
> > second.
> > (To 15 different .rrd files, step 1 sec, see rrdinfo output below)
> > This works fine, (CPU load < 5%, including OS etc) but when I
> > increase the load to 20 updates per sec, (on 20 files), rrdtool
> > does not respond within my 3 sec timeout, after this the rrdtool
> > process hogs the cpu and has to be manually killed.
> Are you *sure* you are only feeding 20 updates per second and,
> for instance, are not in a tight loop?
Yes, there are 20 processes that each sends one update and then
sleeps for one second.
Up to this afternoon it would work with 15 updates and always fail
with 20, but suddenly it started to work with 20 updates/sec,
but fail with 30 updates/sec.
There was one thing wrong in my first mail. After the timeout
rrdtool does not loop forever. when my process that controls the
rrdtool process gets the timeout it terminates, and I assume that
the pipes to the rrdtool process are closed. Normally that would
terminate also the rrdtool process, but it continues running at
100% cpu load for ~5 seconds more before it stops.
> Here's a test I did this afternoon. If it does _not_ handle
> the case where you run into problems, create an example yourself
> and share it. Please, when you do so, try to keep it simple _and_
> do it on the list.
Your test works fine, (of course), I will try to make a simple testcase
that works, but since my system is written in erlang I will start by
investigating the libraries I use to connect to rrdtool.
Unsubscribe mailto:rrd-users-request at list.ee.ethz.ch?subject=unsubscribe
Help mailto:rrd-users-request at list.ee.ethz.ch?subject=help
More information about the rrd-users