It has happened again. Twice today. We have one perf dump. But it is over
600 MB of bzip2-ed data (3.5 GB of plain binary perf data).
We'll go deeper into that, and if we don't come to any conclusions we will
upload it to our anonymous FTP.
Don't know if this perf dump is ok, as it was again triggered by our sciprt
upon system load rise. Maybe it is too late...
Another question: is errors count from my strace log not high?
--
Andrzej Krawiec
2012/6/8 Robert Haas <robertmhaas@gmail.com>
> On Fri, Jun 1, 2012 at 2:17 AM, Andrzej Krawiec
> <a.krawiec@focustelecom.pl> wrote:
> > 2012/5/31 Robert Haas <robertmhaas@gmail.com>:
> >> How long was strace -s run for to generate this?
> >
> > Strace - s was running for about 2 minutes.
>
> Hmm, I'm sort of confused then. This only shows a total of 1.815816
> seconds of system time, which is a lot less than 99% utilization * 2
> minutes.
>
>
