misc/130954: can not obtain which process take all CPU time
Mikolaj Golub
to.my.trociny at gmail.com
Sun Jan 25 02:30:03 PST 2009
The following reply was made to PR bin/130954; it has been noted by GNATS.
From: Mikolaj Golub <to.my.trociny at gmail.com>
To: Eugen Konkov <kes-kes at yandex.ru>
Cc: freebsd-gnats-submit at FreeBSD.org
Subject: Re: misc/130954: can not obtain which process take all CPU time
Date: Sun, 25 Jan 2009 01:56:57 -0800 (PST)
Can you specify how you run gzip? Do you run gzip on one large file or rather
there are many gzip processes spawned one by one?
I can't reproduce this on 7.1-RELEASE-p1 amd64 and 7.0-STABLE (Jun18 2008)
i386 (I don't have 7.1-RELEASE i386 at present).
I have been observing the same on my loaded nagios host (6.3 i386), but I
interpret it in this way. There are many short living processes spawned every
second here, which are born and die faster then top update its statistics, so
the top can't show them.
You can check if it is your case, running something like this in the problem
period:
while sleep 1; do sysctl kern.lastpid; done
If last pid grows fast you have the case I have described above.
If it is not your case and you are observing the problem running one gzip
process on large file, can you check if you see the same running something
like this:
gzip -c - < /dev/random > /dev/null
If you don't, add file system IO:
gzip -c - < /dev/random > test
gzip -c - < test > /dev/null
gzip -c - < test > test1
Also, it would be nice to check if you have your world in sync with kernel.
I think '/usr/bin/top -v', 'ident /usr/bin/top' could be helpful here.
--
Mikolaj Golub
More information about the freebsd-bugs
mailing list