Hi Jack,
Bet you thought we ignored this message. ;-) I'm still working my way
through pre-supercomputing emails.
This was brought up by Nils at the SC PAPI User's Group meeting as well.
It has always been my intention to someday support system level
counters. PAPI uses the term 'granularity' to
To clarify, I saw these problems with Gmond not gmeta. On my stock
RH8.0 gmeta seems to work fine. Gmond's metric-gathering threads die,
and I see only CLUSTER and GANGLIA_XML tags, but no METRICs.
However, if I let the service go on long enough, some metrics get
through occasionally. They end
I have seen this too. I tried debugging it for a while, and got pretty
much the same backtrace. Unfortunately I was sidetracked with other
projects, but this is definately something we need to fix.
One question that has been bothering me is this: how do we have
machines.c files for two differe
And here I was, desperately trying to find out why my modified gmetad
dumped core on me intermittently, when I finally thought "OK, let's
backtrack to the clean 2.5.1 source and see how that behaves". Turned
out that too dies with a segfault. And in libc, no less. Grrr.
Program received signal S
