Hi Ken,
Thanks for your inputs. My host is "xcat20mn". I have only 1 entry in the
"/var/lib/pcp/config/pmlogger/control" file but an interesting thing is
that I have 2 pmlogger process running (I don't know how). I am sending
all the files and o/p you have asked for, including the control file
attached herewith.
[***@xcat20mn ~]# pcp
Performance Co-Pilot configuration on xcat20mn.clusters.com:
platform: Linux xcat20mn 2.6.18-8.el5 #1 SMP Fri Jan 26 14:15:14 EST 2007
x86_64 ia32
hardware: 8 cpus, 2 disks, 6035MB RAM
timezone: EDT+4
licenses: none
pmcd: Version 2.7.4-20080507, 5 agents
pmda: pmcd trace linux apache mounts
[***@xcat20mn ~]#
[***@xcat20mn ~]# pcp -h xcat20mn
Performance Co-Pilot configuration on xcat20mn:
platform: Linux xcat20mn 2.6.18-8.el5 #1 SMP Fri Jan 26 14:15:14 EST 2007
x86_64 ia32
hardware: 8 cpus, 2 disks, 6035MB RAM
timezone: EDT+4
licenses: none
pmcd: Version 2.7.4-20080507, 5 agents
pmda: pmcd trace linux apache mounts
[***@xcat20mn ~]# cat /var/log/pcp/pmcd/pmcd.log
Log for pmcd on xcat20mn started Tue Sep 2 09:00:23 2008
active agent dom pid in out ver protocol parameters
============ === ===== === === === ======== ==========
pmcd 2 2 dso i:2
lib=/var/lib/pcp/pmdas/pmcd/pmda_pmcd.so entry=pmcd_init [0xf7ffa1c4]
linux 60 2 dso i:3
lib=/var/lib/pcp/pmdas/linux/pmda_linux.so entry=linux_init [0xf7fb78eb]
apache 68 31892 10 11 2 bin pipe
cmd=/var/lib/pcp/pmdas/apache/pmdaapache -d 68
mounts 72 31893 12 13 2 bin pipe
cmd=/var/lib/pcp/pmdas/mounts/pmdamounts -d 72
trace 10 31894 14 15 2 bin pipe
cmd=/var/lib/pcp/pmdas/trace/pmdatrace -d 10
Host access list empty: access control turned off
pmcd: PID = 31876, PDU version = 2, pcpcol license
pmcd request port(s):
sts fd port IP addr
=== === ===== ==========
ok 0 44321 0x00000000 INADDR_ANY
ok 3 4321 0x00000000 INADDR_ANY
[***@xcat20mn ~]# ps -ef | grep pmlogger
root 713 25144 0 09:01 pts/11 00:00:00 grep pmlogger
root 31895 31883 0 09:00 pts/11 00:00:00 /bin/sh
/usr/share/pcp/bin/pmlogger_check
root 31913 31895 0 09:00 pts/11 00:00:00 /bin/sh
/usr/share/pcp/bin/pmlogger_check
[***@xcat20mn ~]# pminfo -f pmcd.pmlogger
pmcd.pmlogger.host
No value(s) available!
pmcd.pmlogger.port
No value(s) available!
pmcd.pmlogger.archive
No value(s) available!
pmcd.pmlogger.pmcd_host
No value(s) available!
Thanks and Regards
Sai Prakash
Poughkeepsie Unix Development Lab
IBM Systems and Technology Group
External: 845-435-4720
email: ***@us.ibm.com
Notes: Sai p Seshasayee/Poughkeepsie/IBM
"Always speak the truth, that way you dont need to remember what you said
!"
Ken McDonell <***@internode.on.net>
Sent by: pcp-***@oss.sgi.com
08/29/2008 05:19 PM
Please respond to
***@internode.on.net
To
Sai p Seshasayee/Poughkeepsie/***@IBMUS
cc
Nathan Scott <***@aconex.com>, ***@oss.sgi.com
Subject
Re: PCP - pmlogger command
Sai,
You're trying to start the primary logger here. What is the name of the
host pmlogger is running on?
If it xcat20mn then your pmlogger control file is OK, but there is a
problem with pmcd on this host ... please send output from
$ pcp
$ pcp -h xcat20mn
$ cat /var/log/pcp/pmcd/pmcd.log
If it not, then you need to amend your control file ... change the first
y to n and then run (and send the output from)
# /usr/lib/pcp/bin/pmlogger_check -V
Also (since I suspect you may be trying to start more than one pmlogger
for the same host), it would help to see output from
1. ps -ef | grep pmlogger
2. pminfo -f pmcd.pmlogger
3. and all your pmlogger control file
Hi Nathan,
I set up the pmlogger as a daemon by making the following entry in the
xcat20mn y n PCP_LOG_DIR/pmlogger/xcat20mn -cconfig.base -T1min sai_logger
When I restarted the pcp start scripts, I was able to see the
"pmlogger_check script" running correctly. However, it was not
Subject: pmlogger_check failed in /etc/init.d/pcp
Restarting primary pmlogger for host "xcat20mn" ... timed out
waiting!
Usage: pmlogger [options] archive
-c configfile file to load configuration from
-h host metrics source is PMCD on host
-l logfile redirect diagnostics and trace output
-L linger, even if not primary logger instance and
nothing to log
-n pmnsfile use an alternative PMNS
-P execute as primary logger instance
-r report record sizes and archive growth rate
-s endsize terminate after endsize has been accumulated
-t interval default logging interval [default 60.0
seconds]
-T endtime terminate at given time
-u output is unbuffered
-v volsize switch log volumes after volsize has been
accumulated
-V version generate version 1 or 2 archives (default is 2)
-x fd control file descriptor for application
launching pmlogger
via pmRecordControl(3)
pmlogger_check: Error: cannot find pmlogger output file at
"pmlogger.log"
total 8
drwxr-xr-x 2 root root 4096 Aug 29 12:00 .
drwxr-xr-x 3 root root 4096 Aug 18 10:40 ..
---------- 1 root root 0 Aug 29 12:00 lock
-rw-r--r-- 1 root root 0 Aug 29 11:59 pmlogger.log.prior
pmlogger_check: Error: archive file 20080829.12.00.0 missing
total 8
drwxr-xr-x 2 root root 4096 Aug 29 12:00 .
drwxr-xr-x 3 root root 4096 Aug 18 10:40 ..
---------- 1 root root 0 Aug 29 12:00 lock
-rw-r--r-- 1 root root 0 Aug 29 11:59 pmlogger.log.prior
I tried various combinations in the pmlogger's "control" file entry. I
also created a "pmlogger.log" under "/var/log/pcp/pmlogger/xcat20mn"
but I found that when the daemon finished running, the "pmlogger.log"
file was getting renamed as "pmlogger.log" and gave the above
mentioned output. Any ideas??
Thanks and Regards
Sai Prakash
Poughkeepsie Unix Development Lab
IBM Systems and Technology Group
External: 845-435-4720
Notes: Sai p Seshasayee/Poughkeepsie/IBM
"Always speak the truth, that way you dont need to remember what you
said !"
08/28/2008 07:32 PM
To
Sai p
cc
Subject
Re: PCP -
pmlogger command
Hi Sai,
(please keep CCing the list - theres lots of interested people there
who can also help, and we all learn from each other that way).
Hi Nathan,
Thanks again for your inputs. I was reading the man pages for the
"pmlogger" command and it looks like an exciting option for us to
store archives. We are using Red Had Enterprise Linux 5 and when I
-bash: pmlogger: command not found
Does pmlogger command work on Linux or is it deprecated like the
"pmkstat" command or is it an Irix specific command? (I also read
the
PCPIntro man page and there is no reference of pmlogger being
deprecated). Please let me know if there is an alternate command.
It isn't deprecated it, no. pmkstat may be the only deprecated tool
in the current PCP release, I think. If you use the updated PCP rpm
that I make available (http://oss.sgi.com/~nathans/pcp/rpm/) it is
in fact removed - these rpms are development snapshots of pending code
for the next "official" SGI PCP release. They include alot of fixes
and alot of new features - and they are quite stable (I use them on
all of our production machines here).
$ rpm -ql pcp | grep pmlogger
It lives in this slightly odd place ($PCP_BINADM_DIR
from /etc/pcp.conf)
because it is not often invoked directly, but rather is usually run via
the PCP start scripts (/etc/init.d/pcp) which are run at system bootup.
See the pmlogger_check(1) man page for details on setting pmlogger up as
a daemon.
cheers.
--
Nathan