• 2 replies
  • Latest Post - ‏2012-09-26T07:46:38Z by SystemAdmin
2402 Posts

Pinned topic Documentation of nmon persistent switches for AIX?

‏2011-12-05T21:03:10Z |
I'm currently trying to update our standards for nmon collection on AIX, and in doing so have come across some features that appear to be under-documented.

We'd like to take advantage of the automatic pruning features available when nmon is configured as a persistent local recording via the SMIT panels. However, one wouldn't want to create a standard that has people tabbing through SMIT panels when a script will do the job, so after I failed to find any documentation for these arguments ("-ydays_file", "-a", etc.) I tried to reverse enginner the code that SMIT creates and executes.

SMIT generates a command like the following:

/usr/lpp/perfagent/ add local_nmon /usr/bin/nmon -f -s 300 -ydays_file=1 -a 30 -o /var/perf/daily/ -t -M -^ -d -A -w 4 -l 150 -I 0.1

which in turn seems to add an entry to the inittab, like the following:

nmon:2:once:/usr/bin/nmon -f -s 300 -ydays_file=1 -a 30 -t -K -V -P -M -NN -W -S -^ -O -L -A -w 4 -l 150 -I 0.1 -o /var/perf/daily/ -ypersistent=1 2>&1 >/dev/

null #Start daily recording

Here we see that nmon does appear to have the following arguments, and I don't see these documented anywhere as of yet.

-ydays_file (presumably the number of days stored in a single file)
-ypersistent (presumably the switch adds the entry in the inittab)
-a (presumably the number of days or files to keep)

So, a couple of questions. If anyone has insight they'd care to share I'd be grateful.

1) Does anyone know where to find documentation for these arguments? It's fairly easy to infer how they work, but documentation would be nice. Specifically I'd like to know if these options are supported when calling nmon/topas_nmon "bare" and not through the SMIT panels (see questions 2a and 2b).

2a) The inittab entry is created by as "once", and as far as I can tell there is no entry added to any crontab (seemingly contrary to Nigel's presentation from the Miami conference, although he might have been referring to the WLE reports being scheduled in crontab rather than the data collection). Is the expected behavior of this that it will only attempt to start at boot time, and if data collection is stopped for some reason, unexpectedly or otherwise, that it will only restart after a boot? If so, then a custom cron entry to restart when necessary is probably warranted.

2b) indicates in its comments that it really shouldn't be called from the command line, but should be invoked only from the SMIT panels. If all it's doing in this case is placing the entry in the inittab, then that's fine as I'd have little use for it anyway.

3) Bonus Question: It seems that disk response time ("-d") was left out of the options available from the SMIT topas screen for nmon. Was that intentional? I suppose I should open a PMR for this.
Thanks in advance for any input.

Updated on 2012-09-26T07:46:38Z at 2012-09-26T07:46:38Z by SystemAdmin
  • SystemAdmin
    2402 Posts

    Re: Documentation of nmon persistent switches for AIX?

    I opened a PMR on this issue and have received a response.

    Disk service times (-d option) are indeed not currently available when creating a persistent local nmon recording, however they provided an efix that modifies topas so that it produces disk service times by default. It's unclear if future versions of the SMIT panels will implement selection for this switch, or if will be have the dire warnings against calling it directly removed.

    As to the question of whether any documentation exists for the three switches relevant to persistent recordings, the answer was "No such documentation available ( Am sorry )."

    Not a satisfying answer for a supported element of the OS, but there it is.

    Finally, the persistent local recordings added by SMIT via seem to have no restart mechanism and are indeed added only to the inittab (and not the crontab).
  • SystemAdmin
    2402 Posts

    Re: Documentation of nmon persistent switches for AIX?

    No documentation - that is what we expected to happen when nmon became "topas".

    "Documentation requirements will reach a maximum shortly after the termination of a program."
    Akin's Laws