> I've tried and I'm afraid I cannot reproduce this ... could you please
> send me ...
>
> 1. contents of /var/pcp/config/pmie/control
Attached
> 2. output from: ls -l /var/log/pcp
[root@latte pmlogger]# ls -l /var/log/pcp
lrwxrwxrwx 1 root root 15 Nov 3 13:59 /var/log/pcp ->
/home/logs/pcp/
> 3. output from: pminfo -f pmcd.pmie
Attached
> 4. output from: ps -ef | grep pmie
[root@latte pmlogger]# ps -ef | grep pmie
root 5932 1 0 Nov24 ? 00:00:05 pmie -b -h
latte.harvard.edu -l
root 8945 1 0 21:40 ? 00:00:00 pmie -b -h
latte.harvard.edu -l
root 9115 1 0 21:40 ? 00:00:00 pmie -b -h
latte.harvard.edu -l
root 9319 1 0 21:40 ? 00:00:00 pmie -b -h
latte.harvard.edu -l
root 9383 6229 0 21:41 pts/0 00:00:00 grep pmie
> 5. output from: ls -l /var/tmp/pmie
[root@latte pmlogger]# ls -l /var/tmp/pmie
total 4
-rw-r--r-- 1 root root 8292 Nov 24 23:35 5932
-rw-r--r-- 1 root root 8292 Nov 26 21:40 8945
-rw-r--r-- 1 root root 8292 Nov 26 21:40 9115
-rw-r--r-- 1 root root 8292 Nov 26 21:40 9319
> 6. output from: pmie_check -V
>
[root@latte pmlogger]# /usr/share/pcp/bin/pmie_check -V
Check pmie -h latte.harvard.edu -l
/var/log/pcp/pmie/latte.harvard.edu/pmie.log ...
No current pmie process exists for:
host = latte.harvard.edu
log file = /var/log/pcp/pmie/latte.harvard.edu/pmie.log
Restarting pmie for host "latte.harvard.edu" ...
+ pmie -b -h latte.harvard.edu -l
/var/log/pcp/pmie/latte.harvard.edu/pmie.log -c pmie_config.localhost...
[process 9908] . done
> 7. output from: pminfo -f pmcd.pmie
Attached
> 8. output from: ps -ef | grep pmie
root 5932 1 0 Nov24 ? 00:00:05 pmie -b -h
latte.harvard.edu -l
root 8945 1 0 21:40 ? 00:00:00 pmie -b -h
latte.harvard.edu -l
root 9115 1 0 21:40 ? 00:00:00 pmie -b -h
latte.harvard.edu -l
root 9319 1 0 21:40 ? 00:00:00 pmie -b -h
latte.harvard.edu -l
root 9908 1 0 21:47 ? 00:00:00 pmie -b -h
latte.harvard.edu -l
root 9965 6229 0 21:48 pts/0 00:00:00 grep p
> 9. output from: ls -l /var/tmp/pmie
[root@latte pmlogger]# ls -l /var/tmp/pmie/
total 5
-rw-r--r-- 1 root root 8292 Nov 24 23:35 5932
-rw-r--r-- 1 root root 8292 Nov 26 21:40 8945
-rw-r--r-- 1 root root 8292 Nov 26 21:40 9115
-rw-r--r-- 1 root root 8292 Nov 26 21:40 9319
-rw-r--r-- 1 root root 8292 Nov 26 21:47 9908
> Thanks.
>
>
> On Mon, 26 Nov 2001, Micah Altman wrote:
>
> >
> > Under RH 6.2 and PCP 2.2.1...
> >
> > When /var/log/pcp is a symbolic link, pmie_check fails to detect that a
> > pmie process exists, even when it already running, and starts another
> > one. This leads to an accumulation of pmie processes and eventually to
> > proc-table overflow.
> >
> > The is a patch for pmie_check that eliminates the problem of pmie_check
> > failing to detect run. The fix works for our case, but is probably
> > incorrect -- so please consider it an illustration. The problem seems to
> > be that the symlink makes $logfile <> $p_logfile when they both, in fact,
> > refer to the same file.
> >
> >
> > *** /usr/share/pcp/bin/pmie_check Sun Aug 12 00:44:16 2001
> > --- pmie_check Sun Nov 25 23:50:38 2001
> > ***************
> > *** 1,4 ****
> > ! #! /bin/sh
> > #Tag 0x00010D13
> > #
> > # Copyright (c) 1998-2000 Silicon Graphics, Inc. All Rights Reserved.
> > --- 1,4 ----
> > ! #! /bin/sh
> > #Tag 0x00010D13
> > #
> > # Copyright (c) 1998-2000 Silicon Graphics, Inc. All Rights Reserved.
> > ***************
> > *** 473,479 ****
> > NR == 2 { printf "p_logfile=\"%s\"\n", $0; next }
> > NR == 3 { printf "p_pmcd_host=\"%s\"\n", $0; next }
> > { next }'`
> > ! if [ "$p_logfile" = $logfile -a "$p_pmcd_host" = "$fqdn" ]
> > then
> > pid=$p_id
> > break
> > --- 473,481 ----
> > NR == 2 { printf "p_logfile=\"%s\"\n", $0; next }
> > NR == 3 { printf "p_pmcd_host=\"%s\"\n", $0; next }
> > { next }'`
> > !
> > ! ##if [ "$p_logfile" = $logfile -a "$p_pmcd_host" = "$fqdn" ]
> > ! if [ $logfile -a "$p_pmcd_host" = "$fqdn" ] # fixes problem when
> > /var/log/pcp is a symlink , but not sure if it's correct
> > then
> > pid=$p_id
> > break
> >
> > ______________________________________________________________________________
> > Micah Altman, Ph.D. Associate Director, Harvard-MIT Data
> > Center
> > Director, Virtual Data Center Project, Harvard U. Library
> > Research Associate, Center for Basic Research in the Social Science,
> > Faculty of Arts & Sciences, Harvard University
> > micah_altman@xxxxxxxxxxx http://data.fas.harvard.edu/micah_altman/
> > "Entia non sunt multiplicanda sine necessitate" - Doctor Invincibilis
> > ("Ad indicia spectate." - Micah's corollary)
> > Disclaimer: This is an unofficial, unsigned, document. (Copyright 2001)
> >
> >
>
>
______________________________________________________________________________
Micah Altman, Ph.D. Associate Director, Harvard-MIT Data Center
Director, Virtual Data Center Project, Harvard U. Library
Research Associate, Center for Basic Research in the Social Science,
Faculty of Arts & Sciences, Harvard University
micah_altman@xxxxxxxxxxx http://data.fas.harvard.edu/micah_altman/
"Entia non sunt multiplicanda sine necessitate" - Doctor Invincibilis
("Ad indicia spectate." - Micah's corollary)
Disclaimer: This is an unofficial, unsigned, document. (Copyright 2001)
pcpdebug.zip
Description: Zip archive
|