[Crash-utility] [PATCH 0/4] speed up handling of dumps with many tasks
Greg Thelen
gthelen at google.com
Fri Apr 6 22:03:03 UTC 2018
On Fri, Apr 6, 2018 at 9:10 AM Dave Anderson <anderson at redhat.com> wrote:
> ----- Original Message -----
> > This series decreases crash startup and 'ps' processing time when
handling dumps
> > with many tasks. Prior to the series a 1M task dump took 45m to load
and 45m
> > more to run ps. Once patched, startup+ps time drops below 40 seconds.
> Hi Greg,
> This sounds really good, although I have to say I've never encountered a
dumpfile
> that had such a large task count. I will review and test the patch-set
next week.
Thanks.
> BTW, did you test this on a live system that has such a large task count?
> Like for example, as I recall, the task_exists() checks that you removed
are
> checking for tasks that were found when scanning the lists, but gone by
the
> time it got around to storing them.
I have used these patches on a live machine. 'ps' runs and appears
correct. Though it's a moving target, so difficult to say if every line is
perfect. I don't see any duplicates. And I see a similar set of pids to
what /bin/ps reports. Though /bin/ps consistently reports a subset (~1000
fewer pids than crash reports). For these /bin/ps-mising pids, they are
visible in /proc/$pid and "kill -0 $pid" also works. So I think it's a
either a procfs bug or /bin/ps bug. stracing /bin/ps doesn't show any
syscall errors.
More information about the Crash-utility
mailing list