[Bug 883] Derived metrics are not recordable

[bugzilla-daemon@xxxxxxxxxxx]

http://oss.sgi.com/bugzilla/show_bug.cgi?id=883


Ken McDonell <kenj@xxxxxxxxxxxxxxxx> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |kenj@xxxxxxxxxxxxxxxx




--- Comment #1 from Ken McDonell <kenj@xxxxxxxxxxxxxxxx>  2010-05-06 16:23:08 
CST ---
Dave,

I am surprised, but this seems to be closer to working than expected!

I am assuming that you set $PCP_DERIVED_CONFIG in the environment and then
started pmchart.

Now, everything seems to be going OK once you start recording, but at the END
of recording I see two popup dialogs, one for each of the messages:

Error: derived metric mem.util.ncache_clean: operand: mem.util.cache_clean:
Unknown metric name

Error: derived metric mem.util.mmapped: operand: mem.util.shmem: Unknown metric
name

Does this match what you saw?

Well if you dig a little deeper ...

$ cd ~/.pcp/pmlogger
$ cat 20100507.07.00.59.folio
PCPFolio
Version: 1
# use pmafm(1) to process this PCP archive folio
#
Created: on bozo at Fri May 7 07:01:00 2010
Creator: pmchart /home/kenj/.pcp/pmlogger/20100507.07.00.59.view
#        Host        Basename
Archive:    bozo        /home/kenj/.pcp/pmlogger/bozo/20100507.07.00.59

So the pmafm folio was created OK.

$ cat bozo/20100507.07.00.59.log
Log for pmlogger on bozo started Fri May  7 07:01:00 2010

Config parsed
Starting logger for host "bozo"
Archive basename: /home/kenj/.pcp/pmlogger/bozo/20100507.07.00.59

Group [2 metrics] {
    mem.util.mapped
    mem.util.inactive
} logged every 2 sec: 84 bytes or 3.46 Mbytes/day

And the pmlogger log file indicates no errors.

$ pminfo -a bozo/20100507.07.00.59
Error: derived metric mem.util.ncache_clean: operand: mem.util.cache_clean:
Unknown metric name
Error: derived metric mem.util.mmapped: operand: mem.util.shmem: Unknown metric
name
mem.util.inactive
mem.util.mapped
mem.util.ncache_clean
mem.util.mmapped
pmcd.pmlogger.archive
pmcd.pmlogger.port
pmcd.pmlogger.host

Aha! There is the SAME message we see from pmchart ... but in this context it
is expected because $PCP_DERIVED_CONFIG is still set and we have not included
the operand metrics for the derived metric expressions in the archive.

Let's try that theory.

$ unset PCP_DERIVED_CONFIG
$ pminfo -a bozo/20100507.07.00.59
mem.util.inactive
mem.util.mapped
pmcd.pmlogger.archive
pmcd.pmlogger.port
pmcd.pmlogger.host

And the Error messages go away.  And as a bonus the the archive does contain
logged values for the derived metric!

$ pmval -a bozo/20100507.07.00.59 mem.util.mapped | head -15

metric:    mem.util.mapped
archive:   bozo/20100507.07.00.59
host:      bozo
start:     Fri May  7 07:01:00 2010
end:       Fri May  7 07:02:14 2010
semantics: instantaneous value
units:     Kbyte
samples:   74
interval:  1.00 sec
07:01:00.127  No values available
07:01:01.127  No values available
07:01:02.127                93260
07:01:03.127                93260
07:01:04.127                93260

Now this pminfo invocation would appear to be coming from pmchart, not any of
the PCP libraries supporting "record" mode, so I think this is one for Nathan
to explain.

-- 
Configure bugmail: http://oss.sgi.com/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.