[clug] Q: Linux process accounting?
grail at goldweb.com.au
Thu Sep 1 02:08:08 GMT 2005
I've install the GNU accounting tools to try to take advantage of the
BSD accounting that I've already got compiled into my kernel. After
some period of time, I ran the following command and got a bunch of
output, a selection of which follows:
franklin:/usr/share/doc/acct# sa --list-all-names --sort-sys-user-div-
414 103886.77re 0.23cp 0avio 1374k
3 12490.00re 0.13cp 0avio 10364k
1 0.95re 0.01cp 0avio 2466k frontend
1 22.55re 0.00cp 0avio 830k apt-get
2 1000.27re 0.00cp 0avio 1741k perl
2 0.27re 0.00cp 0avio 938k
10 0.78re 0.01cp 0avio 1795k amavis-
6 2608.53re 0.00cp 0avio 2043k lmtpd
13 3220.93re 0.01cp 0avio 1679k cleanup
12 9479.47re 0.01cp 0avio 1798k smtpd
54 47830.43re 0.02cp 0avio 1639k imapd
As you might guess, this machine is the mail server (SMTP, IMAP)
running Postfix with Amavisd-new configured as the virus/spam catcher.
Can anyone tell me what that first line is? My guess is that it's
either an idle time counter or a total active time counter.
With amavisd-new and imapd consuming so many real seconds for so few
processor seconds, would this imply that it's waiting a long time for
data to arrive through the network or disk? (even though the average
number of I/O calls over the three processes is 0)
Does the prevalence of 0 in the average I/O calls column indicate
that there is something fundamentally wrong with my kernel/accounting
Is there any more documentation out there on how to use the
accounting tools, since all I've found so far is a "HOWTO" on how to
set up the accounting tools (which is no more than "get the tarball,
unpack it, run ./configure, make, make install")?
I really want to know what my machine is doing when I'm not around.
Call me the paranoid parent, but I want it to keep a detailed
personal journal so I can snoop into its private life.
Alex Satrapa M: +61 4 0770 5332
grail at goldweb.com.au W: http://homepage.mac.com/alexsatrapa
More information about the linux